1 USING THE IJG JPEG LIBRARY
3 Copyright (C) 1994-2013, Thomas G. Lane, Guido Vollbeding.
4 This file is part of the Independent JPEG Group's software.
5 For conditions of distribution and use, see the accompanying README file.
8 This file describes how to use the IJG JPEG library within an application
9 program. Read it if you want to write a program that uses the library.
11 The file example.c provides heavily commented skeleton code for calling the
12 JPEG library. Also see jpeglib.h (the include file to be used by application
13 programs) for full details about data structures and function parameter lists.
14 The library source code, of course, is the ultimate reference.
16 Note that there have been *major* changes from the application interface
17 presented by IJG version 4 and earlier versions. The old design had several
18 inherent limitations, and it had accumulated a lot of cruft as we added
19 features while trying to minimize application-interface changes. We have
20 sacrificed backward compatibility in the version 5 rewrite, but we think the
21 improvements justify this.
28 Functions provided by the library
29 Outline of typical usage
34 Mechanics of usage: include files, linking, etc
36 Compression parameter selection
37 Decompression parameter selection
40 Compressed data handling (source and destination managers)
42 Progressive JPEG support
44 Abbreviated datastreams and multiple images
46 Raw (downsampled) image data
47 Really raw data: DCT coefficients
51 Library compile-time options
52 Portability considerations
53 Notes for MS-DOS implementors
55 You should read at least the overview and basic usage sections before trying
56 to program with the library. The sections on advanced features can be read
57 if and when you need them.
63 Functions provided by the library
64 ---------------------------------
66 The IJG JPEG library provides C code to read and write JPEG-compressed image
67 files. The surrounding application program receives or supplies image data a
68 scanline at a time, using a straightforward uncompressed image format. All
69 details of color conversion and other preprocessing/postprocessing can be
70 handled by the library.
72 The library includes a substantial amount of code that is not covered by the
73 JPEG standard but is necessary for typical applications of JPEG. These
74 functions preprocess the image before JPEG compression or postprocess it after
75 decompression. They include colorspace conversion, downsampling/upsampling,
76 and color quantization. The application indirectly selects use of this code
77 by specifying the format in which it wishes to supply or receive image data.
78 For example, if colormapped output is requested, then the decompression
79 library automatically invokes color quantization.
81 A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
82 and even more so in decompression postprocessing. The decompression library
83 provides multiple implementations that cover most of the useful tradeoffs,
84 ranging from very-high-quality down to fast-preview operation. On the
85 compression side we have generally not provided low-quality choices, since
86 compression is normally less time-critical. It should be understood that the
87 low-quality modes may not meet the JPEG standard's accuracy requirements;
88 nonetheless, they are useful for viewers.
90 A word about functions *not* provided by the library. We handle a subset of
91 the ISO JPEG standard; most baseline, extended-sequential, and progressive
92 JPEG processes are supported. (Our subset includes all features now in common
93 use.) Unsupported ISO options include:
94 * Hierarchical storage
97 * Nonintegral subsampling ratios
98 We support both 8- and 12-bit data precision, but this is a compile-time
99 choice rather than a run-time choice; hence it is difficult to use both
100 precisions in a single application.
102 By itself, the library handles only interchange JPEG datastreams --- in
103 particular the widely used JFIF file format. The library can be used by
104 surrounding code to process interchange or abbreviated JPEG datastreams that
105 are embedded in more complex file formats. (For example, this library is
106 used by the free LIBTIFF library to support JPEG compression in TIFF.)
109 Outline of typical usage
110 ------------------------
112 The rough outline of a JPEG compression operation is:
114 Allocate and initialize a JPEG compression object
115 Specify the destination for the compressed data (eg, a file)
116 Set parameters for compression, including image size & colorspace
117 jpeg_start_compress(...);
118 while (scan lines remain to be written)
119 jpeg_write_scanlines(...);
120 jpeg_finish_compress(...);
121 Release the JPEG compression object
123 A JPEG compression object holds parameters and working state for the JPEG
124 library. We make creation/destruction of the object separate from starting
125 or finishing compression of an image; the same object can be re-used for a
126 series of image compression operations. This makes it easy to re-use the
127 same parameter settings for a sequence of images. Re-use of a JPEG object
128 also has important implications for processing abbreviated JPEG datastreams,
131 The image data to be compressed is supplied to jpeg_write_scanlines() from
132 in-memory buffers. If the application is doing file-to-file compression,
133 reading image data from the source file is the application's responsibility.
134 The library emits compressed data by calling a "data destination manager",
135 which typically will write the data into a file; but the application can
136 provide its own destination manager to do something else.
138 Similarly, the rough outline of a JPEG decompression operation is:
140 Allocate and initialize a JPEG decompression object
141 Specify the source of the compressed data (eg, a file)
142 Call jpeg_read_header() to obtain image info
143 Set parameters for decompression
144 jpeg_start_decompress(...);
145 while (scan lines remain to be read)
146 jpeg_read_scanlines(...);
147 jpeg_finish_decompress(...);
148 Release the JPEG decompression object
150 This is comparable to the compression outline except that reading the
151 datastream header is a separate step. This is helpful because information
152 about the image's size, colorspace, etc is available when the application
153 selects decompression parameters. For example, the application can choose an
154 output scaling ratio that will fit the image into the available screen size.
156 The decompression library obtains compressed data by calling a data source
157 manager, which typically will read the data from a file; but other behaviors
158 can be obtained with a custom source manager. Decompressed data is delivered
159 into in-memory buffers passed to jpeg_read_scanlines().
161 It is possible to abort an incomplete compression or decompression operation
162 by calling jpeg_abort(); or, if you do not need to retain the JPEG object,
163 simply release it by calling jpeg_destroy().
165 JPEG compression and decompression objects are two separate struct types.
166 However, they share some common fields, and certain routines such as
167 jpeg_destroy() can work on either type of object.
169 The JPEG library has no static variables: all state is in the compression
170 or decompression object. Therefore it is possible to process multiple
171 compression and decompression operations concurrently, using multiple JPEG
174 Both compression and decompression can be done in an incremental memory-to-
175 memory fashion, if suitable source/destination managers are used. See the
176 section on "I/O suspension" for more details.
185 Before diving into procedural details, it is helpful to understand the
186 image data format that the JPEG library expects or returns.
188 The standard input image format is a rectangular array of pixels, with each
189 pixel having the same number of "component" or "sample" values (color
190 channels). You must specify how many components there are and the colorspace
191 interpretation of the components. Most applications will use RGB data
192 (three components per pixel) or grayscale data (one component per pixel).
193 PLEASE NOTE THAT RGB DATA IS THREE SAMPLES PER PIXEL, GRAYSCALE ONLY ONE.
194 A remarkable number of people manage to miss this, only to find that their
195 programs don't work with grayscale JPEG files.
197 There is no provision for colormapped input. JPEG files are always full-color
198 or full grayscale (or sometimes another colorspace such as CMYK). You can
199 feed in a colormapped image by expanding it to full-color format. However
200 JPEG often doesn't work very well with source data that has been colormapped,
201 because of dithering noise. This is discussed in more detail in the JPEG FAQ
202 and the other references mentioned in the README file.
204 Pixels are stored by scanlines, with each scanline running from left to
205 right. The component values for each pixel are adjacent in the row; for
206 example, R,G,B,R,G,B,R,G,B,... for 24-bit RGB color. Each scanline is an
207 array of data type JSAMPLE --- which is typically "unsigned char", unless
208 you've changed jmorecfg.h. (You can also change the RGB pixel layout, say
209 to B,G,R order, by modifying jmorecfg.h. But see the restrictions listed in
210 that file before doing so.)
212 A 2-D array of pixels is formed by making a list of pointers to the starts of
213 scanlines; so the scanlines need not be physically adjacent in memory. Even
214 if you process just one scanline at a time, you must make a one-element
215 pointer array to conform to this structure. Pointers to JSAMPLE rows are of
216 type JSAMPROW, and the pointer to the pointer array is of type JSAMPARRAY.
218 The library accepts or supplies one or more complete scanlines per call.
219 It is not possible to process part of a row at a time. Scanlines are always
220 processed top-to-bottom. You can process an entire image in one call if you
221 have it all in memory, but usually it's simplest to process one scanline at
224 For best results, source data values should have the precision specified by
225 BITS_IN_JSAMPLE (normally 8 bits). For instance, if you choose to compress
226 data that's only 6 bits/channel, you should left-justify each value in a
227 byte before passing it to the compressor. If you need to compress data
228 that has more than 8 bits/channel, compile with BITS_IN_JSAMPLE = 12.
229 (See "Library compile-time options", later.)
232 The data format returned by the decompressor is the same in all details,
233 except that colormapped output is supported. (Again, a JPEG file is never
234 colormapped. But you can ask the decompressor to perform on-the-fly color
235 quantization to deliver colormapped output.) If you request colormapped
236 output then the returned data array contains a single JSAMPLE per pixel;
237 its value is an index into a color map. The color map is represented as
238 a 2-D JSAMPARRAY in which each row holds the values of one color component,
239 that is, colormap[i][j] is the value of the i'th color component for pixel
240 value (map index) j. Note that since the colormap indexes are stored in
241 JSAMPLEs, the maximum number of colors is limited by the size of JSAMPLE
242 (ie, at most 256 colors for an 8-bit JPEG library).
248 Here we revisit the JPEG compression outline given in the overview.
250 1. Allocate and initialize a JPEG compression object.
252 A JPEG compression object is a "struct jpeg_compress_struct". (It also has
253 a bunch of subsidiary structures which are allocated via malloc(), but the
254 application doesn't control those directly.) This struct can be just a local
255 variable in the calling routine, if a single routine is going to execute the
256 whole JPEG compression sequence. Otherwise it can be static or allocated
259 You will also need a structure representing a JPEG error handler. The part
260 of this that the library cares about is a "struct jpeg_error_mgr". If you
261 are providing your own error handler, you'll typically want to embed the
262 jpeg_error_mgr struct in a larger structure; this is discussed later under
263 "Error handling". For now we'll assume you are just using the default error
264 handler. The default error handler will print JPEG error/warning messages
265 on stderr, and it will call exit() if a fatal error occurs.
267 You must initialize the error handler structure, store a pointer to it into
268 the JPEG object's "err" field, and then call jpeg_create_compress() to
269 initialize the rest of the JPEG object.
271 Typical code for this step, if you are using the default error handler, is
273 struct jpeg_compress_struct cinfo;
274 struct jpeg_error_mgr jerr;
276 cinfo.err = jpeg_std_error(&jerr);
277 jpeg_create_compress(&cinfo);
279 jpeg_create_compress allocates a small amount of memory, so it could fail
280 if you are out of memory. In that case it will exit via the error handler;
281 that's why the error handler must be initialized first.
284 2. Specify the destination for the compressed data (eg, a file).
286 As previously mentioned, the JPEG library delivers compressed data to a
287 "data destination" module. The library includes one data destination
288 module which knows how to write to a stdio stream. You can use your own
289 destination module if you want to do something else, as discussed later.
291 If you use the standard destination module, you must open the target stdio
292 stream beforehand. Typical code for this step looks like:
296 if ((outfile = fopen(filename, "wb")) == NULL) {
297 fprintf(stderr, "can't open %s\n", filename);
300 jpeg_stdio_dest(&cinfo, outfile);
302 where the last line invokes the standard destination module.
304 WARNING: it is critical that the binary compressed data be delivered to the
305 output file unchanged. On non-Unix systems the stdio library may perform
306 newline translation or otherwise corrupt binary data. To suppress this
307 behavior, you may need to use a "b" option to fopen (as shown above), or use
308 setmode() or another routine to put the stdio stream in binary mode. See
309 cjpeg.c and djpeg.c for code that has been found to work on many systems.
311 You can select the data destination after setting other parameters (step 3),
312 if that's more convenient. You may not change the destination between
313 calling jpeg_start_compress() and jpeg_finish_compress().
316 3. Set parameters for compression, including image size & colorspace.
318 You must supply information about the source image by setting the following
319 fields in the JPEG object (cinfo structure):
321 image_width Width of image, in pixels
322 image_height Height of image, in pixels
323 input_components Number of color channels (samples per pixel)
324 in_color_space Color space of source image
326 The image dimensions are, hopefully, obvious. JPEG supports image dimensions
327 of 1 to 64K pixels in either direction. The input color space is typically
328 RGB or grayscale, and input_components is 3 or 1 accordingly. (See "Special
329 color spaces", later, for more info.) The in_color_space field must be
330 assigned one of the J_COLOR_SPACE enum constants, typically JCS_RGB or
333 JPEG has a large number of compression parameters that determine how the
334 image is encoded. Most applications don't need or want to know about all
335 these parameters. You can set all the parameters to reasonable defaults by
336 calling jpeg_set_defaults(); then, if there are particular values you want
337 to change, you can do so after that. The "Compression parameter selection"
338 section tells about all the parameters.
340 You must set in_color_space correctly before calling jpeg_set_defaults(),
341 because the defaults depend on the source image colorspace. However the
342 other three source image parameters need not be valid until you call
343 jpeg_start_compress(). There's no harm in calling jpeg_set_defaults() more
344 than once, if that happens to be convenient.
346 Typical code for a 24-bit RGB source image is
348 cinfo.image_width = Width; /* image width and height, in pixels */
349 cinfo.image_height = Height;
350 cinfo.input_components = 3; /* # of color components per pixel */
351 cinfo.in_color_space = JCS_RGB; /* colorspace of input image */
353 jpeg_set_defaults(&cinfo);
354 /* Make optional parameter settings here */
357 4. jpeg_start_compress(...);
359 After you have established the data destination and set all the necessary
360 source image info and other parameters, call jpeg_start_compress() to begin
361 a compression cycle. This will initialize internal state, allocate working
362 storage, and emit the first few bytes of the JPEG datastream header.
366 jpeg_start_compress(&cinfo, TRUE);
368 The "TRUE" parameter ensures that a complete JPEG interchange datastream
369 will be written. This is appropriate in most cases. If you think you might
370 want to use an abbreviated datastream, read the section on abbreviated
373 Once you have called jpeg_start_compress(), you may not alter any JPEG
374 parameters or other fields of the JPEG object until you have completed
375 the compression cycle.
378 5. while (scan lines remain to be written)
379 jpeg_write_scanlines(...);
381 Now write all the required image data by calling jpeg_write_scanlines()
382 one or more times. You can pass one or more scanlines in each call, up
383 to the total image height. In most applications it is convenient to pass
384 just one or a few scanlines at a time. The expected format for the passed
385 data is discussed under "Data formats", above.
387 Image data should be written in top-to-bottom scanline order. The JPEG spec
388 contains some weasel wording about how top and bottom are application-defined
389 terms (a curious interpretation of the English language...) but if you want
390 your files to be compatible with everyone else's, you WILL use top-to-bottom
391 order. If the source data must be read in bottom-to-top order, you can use
392 the JPEG library's virtual array mechanism to invert the data efficiently.
393 Examples of this can be found in the sample application cjpeg.
395 The library maintains a count of the number of scanlines written so far
396 in the next_scanline field of the JPEG object. Usually you can just use
397 this variable as the loop counter, so that the loop test looks like
398 "while (cinfo.next_scanline < cinfo.image_height)".
400 Code for this step depends heavily on the way that you store the source data.
401 example.c shows the following code for the case of a full-size 2-D source
402 array containing 3-byte RGB pixels:
404 JSAMPROW row_pointer[1]; /* pointer to a single row */
405 int row_stride; /* physical row width in buffer */
407 row_stride = image_width * 3; /* JSAMPLEs per row in image_buffer */
409 while (cinfo.next_scanline < cinfo.image_height) {
410 row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
411 jpeg_write_scanlines(&cinfo, row_pointer, 1);
414 jpeg_write_scanlines() returns the number of scanlines actually written.
415 This will normally be equal to the number passed in, so you can usually
416 ignore the return value. It is different in just two cases:
417 * If you try to write more scanlines than the declared image height,
418 the additional scanlines are ignored.
419 * If you use a suspending data destination manager, output buffer overrun
420 will cause the compressor to return before accepting all the passed lines.
421 This feature is discussed under "I/O suspension", below. The normal
422 stdio destination manager will NOT cause this to happen.
423 In any case, the return value is the same as the change in the value of
427 6. jpeg_finish_compress(...);
429 After all the image data has been written, call jpeg_finish_compress() to
430 complete the compression cycle. This step is ESSENTIAL to ensure that the
431 last bufferload of data is written to the data destination.
432 jpeg_finish_compress() also releases working memory associated with the JPEG
437 jpeg_finish_compress(&cinfo);
439 If using the stdio destination manager, don't forget to close the output
440 stdio stream (if necessary) afterwards.
442 If you have requested a multi-pass operating mode, such as Huffman code
443 optimization, jpeg_finish_compress() will perform the additional passes using
444 data buffered by the first pass. In this case jpeg_finish_compress() may take
445 quite a while to complete. With the default compression parameters, this will
448 It is an error to call jpeg_finish_compress() before writing the necessary
449 total number of scanlines. If you wish to abort compression, call
450 jpeg_abort() as discussed below.
452 After completing a compression cycle, you may dispose of the JPEG object
453 as discussed next, or you may use it to compress another image. In that case
454 return to step 2, 3, or 4 as appropriate. If you do not change the
455 destination manager, the new datastream will be written to the same target.
456 If you do not change any JPEG parameters, the new datastream will be written
457 with the same parameters as before. Note that you can change the input image
458 dimensions freely between cycles, but if you change the input colorspace, you
459 should call jpeg_set_defaults() to adjust for the new colorspace; and then
460 you'll need to repeat all of step 3.
463 7. Release the JPEG compression object.
465 When you are done with a JPEG compression object, destroy it by calling
466 jpeg_destroy_compress(). This will free all subsidiary memory (regardless of
467 the previous state of the object). Or you can call jpeg_destroy(), which
468 works for either compression or decompression objects --- this may be more
469 convenient if you are sharing code between compression and decompression
470 cases. (Actually, these routines are equivalent except for the declared type
471 of the passed pointer. To avoid gripes from ANSI C compilers, jpeg_destroy()
472 should be passed a j_common_ptr.)
474 If you allocated the jpeg_compress_struct structure from malloc(), freeing
475 it is your responsibility --- jpeg_destroy() won't. Ditto for the error
480 jpeg_destroy_compress(&cinfo);
485 If you decide to abort a compression cycle before finishing, you can clean up
486 in either of two ways:
488 * If you don't need the JPEG object any more, just call
489 jpeg_destroy_compress() or jpeg_destroy() to release memory. This is
490 legitimate at any point after calling jpeg_create_compress() --- in fact,
491 it's safe even if jpeg_create_compress() fails.
493 * If you want to re-use the JPEG object, call jpeg_abort_compress(), or call
494 jpeg_abort() which works on both compression and decompression objects.
495 This will return the object to an idle state, releasing any working memory.
496 jpeg_abort() is allowed at any time after successful object creation.
498 Note that cleaning up the data destination, if required, is your
499 responsibility; neither of these routines will call term_destination().
500 (See "Compressed data handling", below, for more about that.)
502 jpeg_destroy() and jpeg_abort() are the only safe calls to make on a JPEG
503 object that has reported an error by calling error_exit (see "Error handling"
504 for more info). The internal state of such an object is likely to be out of
505 whack. Either of these two routines will return the object to a known state.
508 Decompression details
509 ---------------------
511 Here we revisit the JPEG decompression outline given in the overview.
513 1. Allocate and initialize a JPEG decompression object.
515 This is just like initialization for compression, as discussed above,
516 except that the object is a "struct jpeg_decompress_struct" and you
517 call jpeg_create_decompress(). Error handling is exactly the same.
521 struct jpeg_decompress_struct cinfo;
522 struct jpeg_error_mgr jerr;
524 cinfo.err = jpeg_std_error(&jerr);
525 jpeg_create_decompress(&cinfo);
527 (Both here and in the IJG code, we usually use variable name "cinfo" for
528 both compression and decompression objects.)
531 2. Specify the source of the compressed data (eg, a file).
533 As previously mentioned, the JPEG library reads compressed data from a "data
534 source" module. The library includes one data source module which knows how
535 to read from a stdio stream. You can use your own source module if you want
536 to do something else, as discussed later.
538 If you use the standard source module, you must open the source stdio stream
539 beforehand. Typical code for this step looks like:
543 if ((infile = fopen(filename, "rb")) == NULL) {
544 fprintf(stderr, "can't open %s\n", filename);
547 jpeg_stdio_src(&cinfo, infile);
549 where the last line invokes the standard source module.
551 WARNING: it is critical that the binary compressed data be read unchanged.
552 On non-Unix systems the stdio library may perform newline translation or
553 otherwise corrupt binary data. To suppress this behavior, you may need to use
554 a "b" option to fopen (as shown above), or use setmode() or another routine to
555 put the stdio stream in binary mode. See cjpeg.c and djpeg.c for code that
556 has been found to work on many systems.
558 You may not change the data source between calling jpeg_read_header() and
559 jpeg_finish_decompress(). If you wish to read a series of JPEG images from
560 a single source file, you should repeat the jpeg_read_header() to
561 jpeg_finish_decompress() sequence without reinitializing either the JPEG
562 object or the data source module; this prevents buffered input data from
566 3. Call jpeg_read_header() to obtain image info.
568 Typical code for this step is just
570 jpeg_read_header(&cinfo, TRUE);
572 This will read the source datastream header markers, up to the beginning
573 of the compressed data proper. On return, the image dimensions and other
574 info have been stored in the JPEG object. The application may wish to
575 consult this information before selecting decompression parameters.
577 More complex code is necessary if
578 * A suspending data source is used --- in that case jpeg_read_header()
579 may return before it has read all the header data. See "I/O suspension",
580 below. The normal stdio source manager will NOT cause this to happen.
581 * Abbreviated JPEG files are to be processed --- see the section on
582 abbreviated datastreams. Standard applications that deal only in
583 interchange JPEG files need not be concerned with this case either.
585 It is permissible to stop at this point if you just wanted to find out the
586 image dimensions and other header info for a JPEG file. In that case,
587 call jpeg_destroy() when you are done with the JPEG object, or call
588 jpeg_abort() to return it to an idle state before selecting a new data
589 source and reading another header.
592 4. Set parameters for decompression.
594 jpeg_read_header() sets appropriate default decompression parameters based on
595 the properties of the image (in particular, its colorspace). However, you
596 may well want to alter these defaults before beginning the decompression.
597 For example, the default is to produce full color output from a color file.
598 If you want colormapped output you must ask for it. Other options allow the
599 returned image to be scaled and allow various speed/quality tradeoffs to be
600 selected. "Decompression parameter selection", below, gives details.
602 If the defaults are appropriate, nothing need be done at this step.
604 Note that all default values are set by each call to jpeg_read_header().
605 If you reuse a decompression object, you cannot expect your parameter
606 settings to be preserved across cycles, as you can for compression.
607 You must set desired parameter values each time.
610 5. jpeg_start_decompress(...);
612 Once the parameter values are satisfactory, call jpeg_start_decompress() to
613 begin decompression. This will initialize internal state, allocate working
614 memory, and prepare for returning data.
618 jpeg_start_decompress(&cinfo);
620 If you have requested a multi-pass operating mode, such as 2-pass color
621 quantization, jpeg_start_decompress() will do everything needed before data
622 output can begin. In this case jpeg_start_decompress() may take quite a while
623 to complete. With a single-scan (non progressive) JPEG file and default
624 decompression parameters, this will not happen; jpeg_start_decompress() will
627 After this call, the final output image dimensions, including any requested
628 scaling, are available in the JPEG object; so is the selected colormap, if
629 colormapped output has been requested. Useful fields include
631 output_width image width and height, as scaled
633 out_color_components # of color components in out_color_space
634 output_components # of color components returned per pixel
635 colormap the selected colormap, if any
636 actual_number_of_colors number of entries in colormap
638 output_components is 1 (a colormap index) when quantizing colors; otherwise it
639 equals out_color_components. It is the number of JSAMPLE values that will be
640 emitted per pixel in the output arrays.
642 Typically you will need to allocate data buffers to hold the incoming image.
643 You will need output_width * output_components JSAMPLEs per scanline in your
644 output buffer, and a total of output_height scanlines will be returned.
646 Note: if you are using the JPEG library's internal memory manager to allocate
647 data buffers (as djpeg does), then the manager's protocol requires that you
648 request large buffers *before* calling jpeg_start_decompress(). This is a
649 little tricky since the output_XXX fields are not normally valid then. You
650 can make them valid by calling jpeg_calc_output_dimensions() after setting the
651 relevant parameters (scaling, output color space, and quantization flag).
654 6. while (scan lines remain to be read)
655 jpeg_read_scanlines(...);
657 Now you can read the decompressed image data by calling jpeg_read_scanlines()
658 one or more times. At each call, you pass in the maximum number of scanlines
659 to be read (ie, the height of your working buffer); jpeg_read_scanlines()
660 will return up to that many lines. The return value is the number of lines
661 actually read. The format of the returned data is discussed under "Data
662 formats", above. Don't forget that grayscale and color JPEGs will return
663 different data formats!
665 Image data is returned in top-to-bottom scanline order. If you must write
666 out the image in bottom-to-top order, you can use the JPEG library's virtual
667 array mechanism to invert the data efficiently. Examples of this can be
668 found in the sample application djpeg.
670 The library maintains a count of the number of scanlines returned so far
671 in the output_scanline field of the JPEG object. Usually you can just use
672 this variable as the loop counter, so that the loop test looks like
673 "while (cinfo.output_scanline < cinfo.output_height)". (Note that the test
674 should NOT be against image_height, unless you never use scaling. The
675 image_height field is the height of the original unscaled image.)
676 The return value always equals the change in the value of output_scanline.
678 If you don't use a suspending data source, it is safe to assume that
679 jpeg_read_scanlines() reads at least one scanline per call, until the
680 bottom of the image has been reached.
682 If you use a buffer larger than one scanline, it is NOT safe to assume that
683 jpeg_read_scanlines() fills it. (The current implementation returns only a
684 few scanlines per call, no matter how large a buffer you pass.) So you must
685 always provide a loop that calls jpeg_read_scanlines() repeatedly until the
686 whole image has been read.
689 7. jpeg_finish_decompress(...);
691 After all the image data has been read, call jpeg_finish_decompress() to
692 complete the decompression cycle. This causes working memory associated
693 with the JPEG object to be released.
697 jpeg_finish_decompress(&cinfo);
699 If using the stdio source manager, don't forget to close the source stdio
702 It is an error to call jpeg_finish_decompress() before reading the correct
703 total number of scanlines. If you wish to abort decompression, call
704 jpeg_abort() as discussed below.
706 After completing a decompression cycle, you may dispose of the JPEG object as
707 discussed next, or you may use it to decompress another image. In that case
708 return to step 2 or 3 as appropriate. If you do not change the source
709 manager, the next image will be read from the same source.
712 8. Release the JPEG decompression object.
714 When you are done with a JPEG decompression object, destroy it by calling
715 jpeg_destroy_decompress() or jpeg_destroy(). The previous discussion of
716 destroying compression objects applies here too.
720 jpeg_destroy_decompress(&cinfo);
725 You can abort a decompression cycle by calling jpeg_destroy_decompress() or
726 jpeg_destroy() if you don't need the JPEG object any more, or
727 jpeg_abort_decompress() or jpeg_abort() if you want to reuse the object.
728 The previous discussion of aborting compression cycles applies here too.
731 Mechanics of usage: include files, linking, etc
732 -----------------------------------------------
734 Applications using the JPEG library should include the header file jpeglib.h
735 to obtain declarations of data types and routines. Before including
736 jpeglib.h, include system headers that define at least the typedefs FILE and
737 size_t. On ANSI-conforming systems, including <stdio.h> is sufficient; on
738 older Unix systems, you may need <sys/types.h> to define size_t.
740 If the application needs to refer to individual JPEG library error codes, also
741 include jerror.h to define those symbols.
743 jpeglib.h indirectly includes the files jconfig.h and jmorecfg.h. If you are
744 installing the JPEG header files in a system directory, you will want to
745 install all four files: jpeglib.h, jerror.h, jconfig.h, jmorecfg.h.
747 The most convenient way to include the JPEG code into your executable program
748 is to prepare a library file ("libjpeg.a", or a corresponding name on non-Unix
749 machines) and reference it at your link step. If you use only half of the
750 library (only compression or only decompression), only that much code will be
751 included from the library, unless your linker is hopelessly brain-damaged.
752 The supplied makefiles build libjpeg.a automatically (see install.txt).
754 While you can build the JPEG library as a shared library if the whim strikes
755 you, we don't really recommend it. The trouble with shared libraries is that
756 at some point you'll probably try to substitute a new version of the library
757 without recompiling the calling applications. That generally doesn't work
758 because the parameter struct declarations usually change with each new
759 version. In other words, the library's API is *not* guaranteed binary
760 compatible across versions; we only try to ensure source-code compatibility.
761 (In hindsight, it might have been smarter to hide the parameter structs from
762 applications and introduce a ton of access functions instead. Too late now,
765 On some systems your application may need to set up a signal handler to ensure
766 that temporary files are deleted if the program is interrupted. This is most
767 critical if you are on MS-DOS and use the jmemdos.c memory manager back end;
768 it will try to grab extended memory for temp files, and that space will NOT be
769 freed automatically. See cjpeg.c or djpeg.c for an example signal handler.
771 It may be worth pointing out that the core JPEG library does not actually
772 require the stdio library: only the default source/destination managers and
773 error handler need it. You can use the library in a stdio-less environment
774 if you replace those modules and use jmemnobs.c (or another memory manager of
775 your own devising). More info about the minimum system library requirements
776 may be found in jinclude.h.
782 Compression parameter selection
783 -------------------------------
785 This section describes all the optional parameters you can set for JPEG
786 compression, as well as the "helper" routines provided to assist in this
787 task. Proper setting of some parameters requires detailed understanding
788 of the JPEG standard; if you don't know what a parameter is for, it's best
789 not to mess with it! See REFERENCES in the README file for pointers to
790 more info about JPEG.
792 It's a good idea to call jpeg_set_defaults() first, even if you plan to set
793 all the parameters; that way your code is more likely to work with future JPEG
794 libraries that have additional parameters. For the same reason, we recommend
795 you use a helper routine where one is provided, in preference to twiddling
796 cinfo fields directly.
798 The helper routines are:
800 jpeg_set_defaults (j_compress_ptr cinfo)
801 This routine sets all JPEG parameters to reasonable defaults, using
802 only the input image's color space (field in_color_space, which must
803 already be set in cinfo). Many applications will only need to use
804 this routine and perhaps jpeg_set_quality().
806 jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
807 Sets the JPEG file's colorspace (field jpeg_color_space) as specified,
808 and sets other color-space-dependent parameters appropriately. See
809 "Special color spaces", below, before using this. A large number of
810 parameters, including all per-component parameters, are set by this
811 routine; if you want to twiddle individual parameters you should call
812 jpeg_set_colorspace() before rather than after.
814 jpeg_default_colorspace (j_compress_ptr cinfo)
815 Selects an appropriate JPEG colorspace based on cinfo->in_color_space,
816 and calls jpeg_set_colorspace(). This is actually a subroutine of
817 jpeg_set_defaults(). It's broken out in case you want to change
818 just the colorspace-dependent JPEG parameters.
820 jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
821 Constructs JPEG quantization tables appropriate for the indicated
822 quality setting. The quality value is expressed on the 0..100 scale
823 recommended by IJG (cjpeg's "-quality" switch uses this routine).
824 Note that the exact mapping from quality values to tables may change
825 in future IJG releases as more is learned about DCT quantization.
826 If the force_baseline parameter is TRUE, then the quantization table
827 entries are constrained to the range 1..255 for full JPEG baseline
828 compatibility. In the current implementation, this only makes a
829 difference for quality settings below 25, and it effectively prevents
830 very small/low quality files from being generated. The IJG decoder
831 is capable of reading the non-baseline files generated at low quality
832 settings when force_baseline is FALSE, but other decoders may not be.
834 jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
835 boolean force_baseline)
836 Same as jpeg_set_quality() except that the generated tables are the
837 sample tables given in the JPEC spec section K.1, multiplied by the
838 specified scale factor (which is expressed as a percentage; thus
839 scale_factor = 100 reproduces the spec's tables). Note that larger
840 scale factors give lower quality. This entry point is useful for
841 conforming to the Adobe PostScript DCT conventions, but we do not
842 recommend linear scaling as a user-visible quality scale otherwise.
843 force_baseline again constrains the computed table entries to 1..255.
845 int jpeg_quality_scaling (int quality)
846 Converts a value on the IJG-recommended quality scale to a linear
847 scaling percentage. Note that this routine may change or go away
848 in future releases --- IJG may choose to adopt a scaling method that
849 can't be expressed as a simple scalar multiplier, in which case the
850 premise of this routine collapses. Caveat user.
852 jpeg_default_qtables (j_compress_ptr cinfo, boolean force_baseline)
853 Set default quantization tables with linear q_scale_factor[] values
856 jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
857 const unsigned int *basic_table,
858 int scale_factor, boolean force_baseline)
859 Allows an arbitrary quantization table to be created. which_tbl
860 indicates which table slot to fill. basic_table points to an array
861 of 64 unsigned ints given in normal array order. These values are
862 multiplied by scale_factor/100 and then clamped to the range 1..65535
863 (or to 1..255 if force_baseline is TRUE).
864 CAUTION: prior to library version 6a, jpeg_add_quant_table expected
865 the basic table to be given in JPEG zigzag order. If you need to
866 write code that works with either older or newer versions of this
867 routine, you must check the library version number. Something like
868 "#if JPEG_LIB_VERSION >= 61" is the right test.
870 jpeg_simple_progression (j_compress_ptr cinfo)
871 Generates a default scan script for writing a progressive-JPEG file.
872 This is the recommended method of creating a progressive file,
873 unless you want to make a custom scan sequence. You must ensure that
874 the JPEG color space is set correctly before calling this routine.
877 Compression parameters (cinfo fields) include:
880 If TRUE, use arithmetic coding.
881 If FALSE, use Huffman coding.
884 Set DCT block size. All N from 1 to 16 are possible.
885 Default is 8 (baseline format).
886 Larger values produce higher compression,
887 smaller values produce higher quality.
888 An exact DCT stage is possible with 1 or 2.
889 With the default quality of 75 and default Luminance qtable
890 the DCT+Quantization stage is lossless for value 1.
891 Note that values other than 8 require a SmartScale capable decoder,
892 introduced with IJG JPEG 8. Setting the block_size parameter for
893 compression works with version 8c and later.
895 J_DCT_METHOD dct_method
896 Selects the algorithm used for the DCT step. Choices are:
897 JDCT_ISLOW: slow but accurate integer algorithm
898 JDCT_IFAST: faster, less accurate integer method
899 JDCT_FLOAT: floating-point method
900 JDCT_DEFAULT: default method (normally JDCT_ISLOW)
901 JDCT_FASTEST: fastest method (normally JDCT_IFAST)
902 The FLOAT method is very slightly more accurate than the ISLOW method,
903 but may give different results on different machines due to varying
904 roundoff behavior. The integer methods should give the same results
905 on all machines. On machines with sufficiently fast FP hardware, the
906 floating-point method may also be the fastest. The IFAST method is
907 considerably less accurate than the other two; its use is not
908 recommended if high quality is a concern. JDCT_DEFAULT and
909 JDCT_FASTEST are macros configurable by each installation.
911 unsigned int scale_num, scale_denom
912 Scale the image by the fraction scale_num/scale_denom. Default is
913 1/1, or no scaling. Currently, the supported scaling ratios are
914 M/N with all N from 1 to 16, where M is the destination DCT size,
915 which is 8 by default (see block_size parameter above).
916 (The library design allows for arbitrary scaling ratios but this
917 is not likely to be implemented any time soon.)
919 J_COLOR_SPACE jpeg_color_space
921 The JPEG color space and corresponding number of components; see
922 "Special color spaces", below, for more info. We recommend using
923 jpeg_set_colorspace() if you want to change these.
925 J_COLOR_TRANSFORM color_transform
926 Internal color transform identifier, writes LSE marker if nonzero
927 (requires decoder with inverse color transform support, introduced
929 Two values are currently possible: JCT_NONE and JCT_SUBTRACT_GREEN.
930 Set this value for lossless RGB application *before* calling
931 jpeg_set_colorspace(), because entropy table assignment in
932 jpeg_set_colorspace() depends on color_transform.
934 boolean optimize_coding
935 TRUE causes the compressor to compute optimal Huffman coding tables
936 for the image. This requires an extra pass over the data and
937 therefore costs a good deal of space and time. The default is
938 FALSE, which tells the compressor to use the supplied or default
939 Huffman tables. In most cases optimal tables save only a few percent
940 of file size compared to the default tables. Note that when this is
941 TRUE, you need not supply Huffman tables at all, and any you do
942 supply will be overwritten.
944 unsigned int restart_interval
946 To emit restart markers in the JPEG file, set one of these nonzero.
947 Set restart_interval to specify the exact interval in MCU blocks.
948 Set restart_in_rows to specify the interval in MCU rows. (If
949 restart_in_rows is not 0, then restart_interval is set after the
950 image width in MCUs is computed.) Defaults are zero (no restarts).
951 One restart marker per MCU row is often a good choice.
952 NOTE: the overhead of restart markers is higher in grayscale JPEG
953 files than in color files, and MUCH higher in progressive JPEGs.
954 If you use restarts, you may want to use larger intervals in those
957 const jpeg_scan_info * scan_info
959 By default, scan_info is NULL; this causes the compressor to write a
960 single-scan sequential JPEG file. If not NULL, scan_info points to
961 an array of scan definition records of length num_scans. The
962 compressor will then write a JPEG file having one scan for each scan
963 definition record. This is used to generate noninterleaved or
964 progressive JPEG files. The library checks that the scan array
965 defines a valid JPEG scan sequence. (jpeg_simple_progression creates
966 a suitable scan definition array for progressive JPEG.) This is
967 discussed further under "Progressive JPEG support".
969 boolean do_fancy_downsampling
970 If TRUE, use direct DCT scaling with DCT size > 8 for downsampling
971 of chroma components.
972 If FALSE, use only DCT size <= 8 and simple separate downsampling.
974 For better image stability in multiple generation compression cycles
975 it is preferable that this value matches the corresponding
976 do_fancy_upsampling value in decompression.
979 If non-zero, the input image is smoothed; the value should be 1 for
980 minimal smoothing to 100 for maximum smoothing. Consult jcsample.c
981 for details of the smoothing algorithm. The default is zero.
983 boolean write_JFIF_header
984 If TRUE, a JFIF APP0 marker is emitted. jpeg_set_defaults() and
985 jpeg_set_colorspace() set this TRUE if a JFIF-legal JPEG color space
986 (ie, YCbCr or grayscale) is selected, otherwise FALSE.
988 UINT8 JFIF_major_version
989 UINT8 JFIF_minor_version
990 The version number to be written into the JFIF marker.
991 jpeg_set_defaults() initializes the version to 1.01 (major=minor=1).
992 You should set it to 1.02 (major=1, minor=2) if you plan to write
993 any JFIF 1.02 extension markers.
998 The resolution information to be written into the JFIF marker;
999 not used otherwise. density_unit may be 0 for unknown,
1000 1 for dots/inch, or 2 for dots/cm. The default values are 0,1,1
1001 indicating square pixels of unknown size.
1003 boolean write_Adobe_marker
1004 If TRUE, an Adobe APP14 marker is emitted. jpeg_set_defaults() and
1005 jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
1006 or YCCK is selected, otherwise FALSE. It is generally a bad idea
1007 to set both write_JFIF_header and write_Adobe_marker. In fact,
1008 you probably shouldn't change the default settings at all --- the
1009 default behavior ensures that the JPEG file's color space can be
1010 recognized by the decoder.
1012 JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS]
1013 Pointers to coefficient quantization tables, one per table slot,
1014 or NULL if no table is defined for a slot. Usually these should
1015 be set via one of the above helper routines; jpeg_add_quant_table()
1016 is general enough to define any quantization table. The other
1017 routines will set up table slot 0 for luminance quality and table
1018 slot 1 for chrominance.
1020 int q_scale_factor[NUM_QUANT_TBLS]
1021 Linear quantization scaling factors (percentage, initialized 100)
1022 for use with jpeg_default_qtables().
1023 See rdswitch.c and cjpeg.c for an example of usage.
1024 Note that the q_scale_factor[] fields are the "linear" scales, so you
1025 have to convert from user-defined ratings via jpeg_quality_scaling().
1026 Here is an example code which corresponds to cjpeg -quality 90,70:
1028 jpeg_set_defaults(cinfo);
1030 /* Set luminance quality 90. */
1031 cinfo->q_scale_factor[0] = jpeg_quality_scaling(90);
1032 /* Set chrominance quality 70. */
1033 cinfo->q_scale_factor[1] = jpeg_quality_scaling(70);
1035 jpeg_default_qtables(cinfo, force_baseline);
1037 CAUTION: You must also set 1x1 subsampling for efficient separate
1038 color quality selection, since the default value used by library
1041 cinfo->comp_info[0].v_samp_factor = 1;
1042 cinfo->comp_info[0].h_samp_factor = 1;
1044 JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS]
1045 JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS]
1046 Pointers to Huffman coding tables, one per table slot, or NULL if
1047 no table is defined for a slot. Slots 0 and 1 are filled with the
1048 JPEG sample tables by jpeg_set_defaults(). If you need to allocate
1049 more table structures, jpeg_alloc_huff_table() may be used.
1050 Note that optimal Huffman tables can be computed for an image
1051 by setting optimize_coding, as discussed above; there's seldom
1052 any need to mess with providing your own Huffman tables.
1055 The actual dimensions of the JPEG image that will be written to the file are
1056 given by the following fields. These are computed from the input image
1057 dimensions and the compression parameters by jpeg_start_compress(). You can
1058 also call jpeg_calc_jpeg_dimensions() to obtain the values that will result
1059 from the current parameter settings. This can be useful if you are trying
1060 to pick a scaling ratio that will get close to a desired target size.
1062 JDIMENSION jpeg_width Actual dimensions of output image.
1063 JDIMENSION jpeg_height
1066 Per-component parameters are stored in the struct cinfo.comp_info[i] for
1067 component number i. Note that components here refer to components of the
1068 JPEG color space, *not* the source image color space. A suitably large
1069 comp_info[] array is allocated by jpeg_set_defaults(); if you choose not
1070 to use that routine, it's up to you to allocate the array.
1073 The one-byte identifier code to be recorded in the JPEG file for
1074 this component. For the standard color spaces, we recommend you
1075 leave the default values alone.
1079 Horizontal and vertical sampling factors for the component; must
1080 be 1..4 according to the JPEG standard. Note that larger sampling
1081 factors indicate a higher-resolution component; many people find
1082 this behavior quite unintuitive. The default values are 2,2 for
1083 luminance components and 1,1 for chrominance components, except
1084 for grayscale where 1,1 is used.
1087 Quantization table number for component. The default value is
1088 0 for luminance components and 1 for chrominance components.
1092 DC and AC entropy coding table numbers. The default values are
1093 0 for luminance components and 1 for chrominance components.
1096 Must equal the component's index in comp_info[]. (Beginning in
1097 release v6, the compressor library will fill this in automatically;
1101 Decompression parameter selection
1102 ---------------------------------
1104 Decompression parameter selection is somewhat simpler than compression
1105 parameter selection, since all of the JPEG internal parameters are
1106 recorded in the source file and need not be supplied by the application.
1107 (Unless you are working with abbreviated files, in which case see
1108 "Abbreviated datastreams", below.) Decompression parameters control
1109 the postprocessing done on the image to deliver it in a format suitable
1110 for the application's use. Many of the parameters control speed/quality
1111 tradeoffs, in which faster decompression may be obtained at the price of
1112 a poorer-quality image. The defaults select the highest quality (slowest)
1115 The following fields in the JPEG object are set by jpeg_read_header() and
1116 may be useful to the application in choosing decompression parameters:
1118 JDIMENSION image_width Width and height of image
1119 JDIMENSION image_height
1120 int num_components Number of color components
1121 J_COLOR_SPACE jpeg_color_space Colorspace of image
1122 boolean saw_JFIF_marker TRUE if a JFIF APP0 marker was seen
1123 UINT8 JFIF_major_version Version information from JFIF marker
1124 UINT8 JFIF_minor_version
1125 UINT8 density_unit Resolution data from JFIF marker
1128 boolean saw_Adobe_marker TRUE if an Adobe APP14 marker was seen
1129 UINT8 Adobe_transform Color transform code from Adobe marker
1131 The JPEG color space, unfortunately, is something of a guess since the JPEG
1132 standard proper does not provide a way to record it. In practice most files
1133 adhere to the JFIF or Adobe conventions, and the decoder will recognize these
1134 correctly. See "Special color spaces", below, for more info.
1137 The decompression parameters that determine the basic properties of the
1140 J_COLOR_SPACE out_color_space
1141 Output color space. jpeg_read_header() sets an appropriate default
1142 based on jpeg_color_space; typically it will be RGB or grayscale.
1143 The application can change this field to request output in a different
1144 colorspace. For example, set it to JCS_GRAYSCALE to get grayscale
1145 output from a color file. (This is useful for previewing: grayscale
1146 output is faster than full color since the color components need not
1147 be processed.) Note that not all possible color space transforms are
1148 currently implemented; you may need to extend jdcolor.c if you want an
1151 unsigned int scale_num, scale_denom
1152 Scale the image by the fraction scale_num/scale_denom. Currently,
1153 the supported scaling ratios are M/N with all M from 1 to 16, where
1154 N is the source DCT size, which is 8 for baseline JPEG. (The library
1155 design allows for arbitrary scaling ratios but this is not likely
1156 to be implemented any time soon.) The values are initialized by
1157 jpeg_read_header() with the source DCT size. For baseline JPEG
1158 this is 8/8. If you change only the scale_num value while leaving
1159 the other unchanged, then this specifies the DCT scaled size to be
1160 applied on the given input. For baseline JPEG this is equivalent
1161 to M/8 scaling, since the source DCT size for baseline JPEG is 8.
1162 Smaller scaling ratios permit significantly faster decoding since
1163 fewer pixels need be processed and a simpler IDCT method can be used.
1165 boolean quantize_colors
1166 If set TRUE, colormapped output will be delivered. Default is FALSE,
1167 meaning that full-color output will be delivered.
1169 The next three parameters are relevant only if quantize_colors is TRUE.
1171 int desired_number_of_colors
1172 Maximum number of colors to use in generating a library-supplied color
1173 map (the actual number of colors is returned in a different field).
1174 Default 256. Ignored when the application supplies its own color map.
1176 boolean two_pass_quantize
1177 If TRUE, an extra pass over the image is made to select a custom color
1178 map for the image. This usually looks a lot better than the one-size-
1179 fits-all colormap that is used otherwise. Default is TRUE. Ignored
1180 when the application supplies its own color map.
1182 J_DITHER_MODE dither_mode
1183 Selects color dithering method. Supported values are:
1184 JDITHER_NONE no dithering: fast, very low quality
1185 JDITHER_ORDERED ordered dither: moderate speed and quality
1186 JDITHER_FS Floyd-Steinberg dither: slow, high quality
1187 Default is JDITHER_FS. (At present, ordered dither is implemented
1188 only in the single-pass, standard-colormap case. If you ask for
1189 ordered dither when two_pass_quantize is TRUE or when you supply
1190 an external color map, you'll get F-S dithering.)
1192 When quantize_colors is TRUE, the target color map is described by the next
1193 two fields. colormap is set to NULL by jpeg_read_header(). The application
1194 can supply a color map by setting colormap non-NULL and setting
1195 actual_number_of_colors to the map size. Otherwise, jpeg_start_decompress()
1196 selects a suitable color map and sets these two fields itself.
1197 [Implementation restriction: at present, an externally supplied colormap is
1198 only accepted for 3-component output color spaces.]
1201 The color map, represented as a 2-D pixel array of out_color_components
1202 rows and actual_number_of_colors columns. Ignored if not quantizing.
1203 CAUTION: if the JPEG library creates its own colormap, the storage
1204 pointed to by this field is released by jpeg_finish_decompress().
1205 Copy the colormap somewhere else first, if you want to save it.
1207 int actual_number_of_colors
1208 The number of colors in the color map.
1210 Additional decompression parameters that the application may set include:
1212 J_DCT_METHOD dct_method
1213 Selects the algorithm used for the DCT step. Choices are the same
1214 as described above for compression.
1216 boolean do_fancy_upsampling
1217 If TRUE, use direct DCT scaling with DCT size > 8 for upsampling
1218 of chroma components.
1219 If FALSE, use only DCT size <= 8 and simple separate upsampling.
1221 For better image stability in multiple generation compression cycles
1222 it is preferable that this value matches the corresponding
1223 do_fancy_downsampling value in compression.
1225 boolean do_block_smoothing
1226 If TRUE, interblock smoothing is applied in early stages of decoding
1227 progressive JPEG files; if FALSE, not. Default is TRUE. Early
1228 progression stages look "fuzzy" with smoothing, "blocky" without.
1229 In any case, block smoothing ceases to be applied after the first few
1230 AC coefficients are known to full accuracy, so it is relevant only
1231 when using buffered-image mode for progressive images.
1233 boolean enable_1pass_quant
1234 boolean enable_external_quant
1235 boolean enable_2pass_quant
1236 These are significant only in buffered-image mode, which is
1237 described in its own section below.
1240 The output image dimensions are given by the following fields. These are
1241 computed from the source image dimensions and the decompression parameters
1242 by jpeg_start_decompress(). You can also call jpeg_calc_output_dimensions()
1243 to obtain the values that will result from the current parameter settings.
1244 This can be useful if you are trying to pick a scaling ratio that will get
1245 close to a desired target size. It's also important if you are using the
1246 JPEG library's memory manager to allocate output buffer space, because you
1247 are supposed to request such buffers *before* jpeg_start_decompress().
1249 JDIMENSION output_width Actual dimensions of output image.
1250 JDIMENSION output_height
1251 int out_color_components Number of color components in out_color_space.
1252 int output_components Number of color components returned.
1253 int rec_outbuf_height Recommended height of scanline buffer.
1255 When quantizing colors, output_components is 1, indicating a single color map
1256 index per pixel. Otherwise it equals out_color_components. The output arrays
1257 are required to be output_width * output_components JSAMPLEs wide.
1259 rec_outbuf_height is the recommended minimum height (in scanlines) of the
1260 buffer passed to jpeg_read_scanlines(). If the buffer is smaller, the
1261 library will still work, but time will be wasted due to unnecessary data
1262 copying. In high-quality modes, rec_outbuf_height is always 1, but some
1263 faster, lower-quality modes set it to larger values (typically 2 to 4).
1264 If you are going to ask for a high-speed processing mode, you may as well
1265 go to the trouble of honoring rec_outbuf_height so as to avoid data copying.
1266 (An output buffer larger than rec_outbuf_height lines is OK, but won't
1267 provide any material speed improvement over that height.)
1270 Special color spaces
1271 --------------------
1273 The JPEG standard itself is "color blind" and doesn't specify any particular
1274 color space. It is customary to convert color data to a luminance/chrominance
1275 color space before compressing, since this permits greater compression. The
1276 existing de-facto JPEG file format standards specify YCbCr or grayscale data
1277 (JFIF), or grayscale, RGB, YCbCr, CMYK, or YCCK (Adobe). For special
1278 applications such as multispectral images, other color spaces can be used,
1279 but it must be understood that such files will be unportable.
1281 The JPEG library can handle the most common colorspace conversions (namely
1282 RGB <=> YCbCr and CMYK <=> YCCK). It can also deal with data of an unknown
1283 color space, passing it through without conversion. If you deal extensively
1284 with an unusual color space, you can easily extend the library to understand
1285 additional color spaces and perform appropriate conversions.
1287 For compression, the source data's color space is specified by field
1288 in_color_space. This is transformed to the JPEG file's color space given
1289 by jpeg_color_space. jpeg_set_defaults() chooses a reasonable JPEG color
1290 space depending on in_color_space, but you can override this by calling
1291 jpeg_set_colorspace(). Of course you must select a supported transformation.
1292 jccolor.c currently supports the following transformations:
1297 plus the null transforms: GRAYSCALE => GRAYSCALE, RGB => RGB,
1298 YCbCr => YCbCr, CMYK => CMYK, YCCK => YCCK, and UNKNOWN => UNKNOWN.
1300 The de-facto file format standards (JFIF and Adobe) specify APPn markers that
1301 indicate the color space of the JPEG file. It is important to ensure that
1302 these are written correctly, or omitted if the JPEG file's color space is not
1303 one of the ones supported by the de-facto standards. jpeg_set_colorspace()
1304 will set the compression parameters to include or omit the APPn markers
1305 properly, so long as it is told the truth about the JPEG color space.
1306 For example, if you are writing some random 3-component color space without
1307 conversion, don't try to fake out the library by setting in_color_space and
1308 jpeg_color_space to JCS_YCbCr; use JCS_UNKNOWN. You may want to write an
1309 APPn marker of your own devising to identify the colorspace --- see "Special
1312 When told that the color space is UNKNOWN, the library will default to using
1313 luminance-quality compression parameters for all color components. You may
1314 well want to change these parameters. See the source code for
1315 jpeg_set_colorspace(), in jcparam.c, for details.
1317 For decompression, the JPEG file's color space is given in jpeg_color_space,
1318 and this is transformed to the output color space out_color_space.
1319 jpeg_read_header's setting of jpeg_color_space can be relied on if the file
1320 conforms to JFIF or Adobe conventions, but otherwise it is no better than a
1321 guess. If you know the JPEG file's color space for certain, you can override
1322 jpeg_read_header's guess by setting jpeg_color_space. jpeg_read_header also
1323 selects a default output color space based on (its guess of) jpeg_color_space;
1324 set out_color_space to override this. Again, you must select a supported
1325 transformation. jdcolor.c currently supports
1331 as well as the null transforms. (Since GRAYSCALE=>RGB is provided, an
1332 application can force grayscale JPEGs to look like color JPEGs if it only
1333 wants to handle one case.)
1335 The two-pass color quantizer, jquant2.c, is specialized to handle RGB data
1336 (it weights distances appropriately for RGB colors). You'll need to modify
1337 the code if you want to use it for non-RGB output color spaces. Note that
1338 jquant2.c is used to map to an application-supplied colormap as well as for
1339 the normal two-pass colormap selection process.
1341 CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
1342 files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
1343 This is arguably a bug in Photoshop, but if you need to work with Photoshop
1344 CMYK files, you will have to deal with it in your application. We cannot
1345 "fix" this in the library by inverting the data during the CMYK<=>YCCK
1346 transform, because that would break other applications, notably Ghostscript.
1347 Photoshop versions prior to 3.0 write EPS files containing JPEG-encoded CMYK
1348 data in the same inverted-YCCK representation used in bare JPEG files, but
1349 the surrounding PostScript code performs an inversion using the PS image
1350 operator. I am told that Photoshop 3.0 will write uninverted YCCK in
1351 EPS/JPEG files, and will omit the PS-level inversion. (But the data
1352 polarity used in bare JPEG files will not change in 3.0.) In either case,
1353 the JPEG library must not invert the data itself, or else Ghostscript would
1354 read these EPS files incorrectly.
1360 When the default error handler is used, any error detected inside the JPEG
1361 routines will cause a message to be printed on stderr, followed by exit().
1362 You can supply your own error handling routines to override this behavior
1363 and to control the treatment of nonfatal warnings and trace/debug messages.
1364 The file example.c illustrates the most common case, which is to have the
1365 application regain control after an error rather than exiting.
1367 The JPEG library never writes any message directly; it always goes through
1368 the error handling routines. Three classes of messages are recognized:
1369 * Fatal errors: the library cannot continue.
1370 * Warnings: the library can continue, but the data is corrupt, and a
1371 damaged output image is likely to result.
1372 * Trace/informational messages. These come with a trace level indicating
1373 the importance of the message; you can control the verbosity of the
1374 program by adjusting the maximum trace level that will be displayed.
1376 You may, if you wish, simply replace the entire JPEG error handling module
1377 (jerror.c) with your own code. However, you can avoid code duplication by
1378 only replacing some of the routines depending on the behavior you need.
1379 This is accomplished by calling jpeg_std_error() as usual, but then overriding
1380 some of the method pointers in the jpeg_error_mgr struct, as illustrated by
1383 All of the error handling routines will receive a pointer to the JPEG object
1384 (a j_common_ptr which points to either a jpeg_compress_struct or a
1385 jpeg_decompress_struct; if you need to tell which, test the is_decompressor
1386 field). This struct includes a pointer to the error manager struct in its
1387 "err" field. Frequently, custom error handler routines will need to access
1388 additional data which is not known to the JPEG library or the standard error
1389 handler. The most convenient way to do this is to embed either the JPEG
1390 object or the jpeg_error_mgr struct in a larger structure that contains
1391 additional fields; then casting the passed pointer provides access to the
1392 additional fields. Again, see example.c for one way to do it. (Beginning
1393 with IJG version 6b, there is also a void pointer "client_data" in each
1394 JPEG object, which the application can also use to find related data.
1395 The library does not touch client_data at all.)
1397 The individual methods that you might wish to override are:
1399 error_exit (j_common_ptr cinfo)
1400 Receives control for a fatal error. Information sufficient to
1401 generate the error message has been stored in cinfo->err; call
1402 output_message to display it. Control must NOT return to the caller;
1403 generally this routine will exit() or longjmp() somewhere.
1404 Typically you would override this routine to get rid of the exit()
1405 default behavior. Note that if you continue processing, you should
1406 clean up the JPEG object with jpeg_abort() or jpeg_destroy().
1408 output_message (j_common_ptr cinfo)
1409 Actual output of any JPEG message. Override this to send messages
1410 somewhere other than stderr. Note that this method does not know
1411 how to generate a message, only where to send it.
1413 format_message (j_common_ptr cinfo, char * buffer)
1414 Constructs a readable error message string based on the error info
1415 stored in cinfo->err. This method is called by output_message. Few
1416 applications should need to override this method. One possible
1417 reason for doing so is to implement dynamic switching of error message
1420 emit_message (j_common_ptr cinfo, int msg_level)
1421 Decide whether or not to emit a warning or trace message; if so,
1422 calls output_message. The main reason for overriding this method
1423 would be to abort on warnings. msg_level is -1 for warnings,
1424 0 and up for trace messages.
1426 Only error_exit() and emit_message() are called from the rest of the JPEG
1427 library; the other two are internal to the error handler.
1429 The actual message texts are stored in an array of strings which is pointed to
1430 by the field err->jpeg_message_table. The messages are numbered from 0 to
1431 err->last_jpeg_message, and it is these code numbers that are used in the
1432 JPEG library code. You could replace the message texts (for instance, with
1433 messages in French or German) by changing the message table pointer. See
1434 jerror.h for the default texts. CAUTION: this table will almost certainly
1435 change or grow from one library version to the next.
1437 It may be useful for an application to add its own message texts that are
1438 handled by the same mechanism. The error handler supports a second "add-on"
1439 message table for this purpose. To define an addon table, set the pointer
1440 err->addon_message_table and the message numbers err->first_addon_message and
1441 err->last_addon_message. If you number the addon messages beginning at 1000
1442 or so, you won't have to worry about conflicts with the library's built-in
1443 messages. See the sample applications cjpeg/djpeg for an example of using
1444 addon messages (the addon messages are defined in cderror.h).
1446 Actual invocation of the error handler is done via macros defined in jerror.h:
1447 ERREXITn(...) for fatal errors
1448 WARNMSn(...) for corrupt-data warnings
1449 TRACEMSn(...) for trace and informational messages.
1450 These macros store the message code and any additional parameters into the
1451 error handler struct, then invoke the error_exit() or emit_message() method.
1452 The variants of each macro are for varying numbers of additional parameters.
1453 The additional parameters are inserted into the generated message using
1454 standard printf() format codes.
1456 See jerror.h and jerror.c for further details.
1459 Compressed data handling (source and destination managers)
1460 ----------------------------------------------------------
1462 The JPEG compression library sends its compressed data to a "destination
1463 manager" module. The default destination manager just writes the data to a
1464 memory buffer or to a stdio stream, but you can provide your own manager to
1465 do something else. Similarly, the decompression library calls a "source
1466 manager" to obtain the compressed data; you can provide your own source
1467 manager if you want the data to come from somewhere other than a memory
1468 buffer or a stdio stream.
1470 In both cases, compressed data is processed a bufferload at a time: the
1471 destination or source manager provides a work buffer, and the library invokes
1472 the manager only when the buffer is filled or emptied. (You could define a
1473 one-character buffer to force the manager to be invoked for each byte, but
1474 that would be rather inefficient.) The buffer's size and location are
1475 controlled by the manager, not by the library. For example, the memory
1476 source manager just makes the buffer pointer and length point to the original
1477 data in memory. In this case the buffer-reload procedure will be invoked
1478 only if the decompressor ran off the end of the datastream, which would
1479 indicate an erroneous datastream.
1481 The work buffer is defined as an array of datatype JOCTET, which is generally
1482 "char" or "unsigned char". On a machine where char is not exactly 8 bits
1483 wide, you must define JOCTET as a wider data type and then modify the data
1484 source and destination modules to transcribe the work arrays into 8-bit units
1485 on external storage.
1487 A data destination manager struct contains a pointer and count defining the
1488 next byte to write in the work buffer and the remaining free space:
1490 JOCTET * next_output_byte; /* => next byte to write in buffer */
1491 size_t free_in_buffer; /* # of byte spaces remaining in buffer */
1493 The library increments the pointer and decrements the count until the buffer
1494 is filled. The manager's empty_output_buffer method must reset the pointer
1495 and count. The manager is expected to remember the buffer's starting address
1496 and total size in private fields not visible to the library.
1498 A data destination manager provides three methods:
1500 init_destination (j_compress_ptr cinfo)
1501 Initialize destination. This is called by jpeg_start_compress()
1502 before any data is actually written. It must initialize
1503 next_output_byte and free_in_buffer. free_in_buffer must be
1504 initialized to a positive value.
1506 empty_output_buffer (j_compress_ptr cinfo)
1507 This is called whenever the buffer has filled (free_in_buffer
1508 reaches zero). In typical applications, it should write out the
1509 *entire* buffer (use the saved start address and buffer length;
1510 ignore the current state of next_output_byte and free_in_buffer).
1511 Then reset the pointer & count to the start of the buffer, and
1512 return TRUE indicating that the buffer has been dumped.
1513 free_in_buffer must be set to a positive value when TRUE is
1514 returned. A FALSE return should only be used when I/O suspension is
1515 desired (this operating mode is discussed in the next section).
1517 term_destination (j_compress_ptr cinfo)
1518 Terminate destination --- called by jpeg_finish_compress() after all
1519 data has been written. In most applications, this must flush any
1520 data remaining in the buffer. Use either next_output_byte or
1521 free_in_buffer to determine how much data is in the buffer.
1523 term_destination() is NOT called by jpeg_abort() or jpeg_destroy(). If you
1524 want the destination manager to be cleaned up during an abort, you must do it
1527 You will also need code to create a jpeg_destination_mgr struct, fill in its
1528 method pointers, and insert a pointer to the struct into the "dest" field of
1529 the JPEG compression object. This can be done in-line in your setup code if
1530 you like, but it's probably cleaner to provide a separate routine similar to
1531 the jpeg_stdio_dest() or jpeg_mem_dest() routines of the supplied destination
1534 Decompression source managers follow a parallel design, but with some
1535 additional frammishes. The source manager struct contains a pointer and count
1536 defining the next byte to read from the work buffer and the number of bytes
1539 const JOCTET * next_input_byte; /* => next byte to read from buffer */
1540 size_t bytes_in_buffer; /* # of bytes remaining in buffer */
1542 The library increments the pointer and decrements the count until the buffer
1543 is emptied. The manager's fill_input_buffer method must reset the pointer and
1544 count. In most applications, the manager must remember the buffer's starting
1545 address and total size in private fields not visible to the library.
1547 A data source manager provides five methods:
1549 init_source (j_decompress_ptr cinfo)
1550 Initialize source. This is called by jpeg_read_header() before any
1551 data is actually read. Unlike init_destination(), it may leave
1552 bytes_in_buffer set to 0 (in which case a fill_input_buffer() call
1553 will occur immediately).
1555 fill_input_buffer (j_decompress_ptr cinfo)
1556 This is called whenever bytes_in_buffer has reached zero and more
1557 data is wanted. In typical applications, it should read fresh data
1558 into the buffer (ignoring the current state of next_input_byte and
1559 bytes_in_buffer), reset the pointer & count to the start of the
1560 buffer, and return TRUE indicating that the buffer has been reloaded.
1561 It is not necessary to fill the buffer entirely, only to obtain at
1562 least one more byte. bytes_in_buffer MUST be set to a positive value
1563 if TRUE is returned. A FALSE return should only be used when I/O
1564 suspension is desired (this mode is discussed in the next section).
1566 skip_input_data (j_decompress_ptr cinfo, long num_bytes)
1567 Skip num_bytes worth of data. The buffer pointer and count should
1568 be advanced over num_bytes input bytes, refilling the buffer as
1569 needed. This is used to skip over a potentially large amount of
1570 uninteresting data (such as an APPn marker). In some applications
1571 it may be possible to optimize away the reading of the skipped data,
1572 but it's not clear that being smart is worth much trouble; large
1573 skips are uncommon. bytes_in_buffer may be zero on return.
1574 A zero or negative skip count should be treated as a no-op.
1576 resync_to_restart (j_decompress_ptr cinfo, int desired)
1577 This routine is called only when the decompressor has failed to find
1578 a restart (RSTn) marker where one is expected. Its mission is to
1579 find a suitable point for resuming decompression. For most
1580 applications, we recommend that you just use the default resync
1581 procedure, jpeg_resync_to_restart(). However, if you are able to back
1582 up in the input data stream, or if you have a-priori knowledge about
1583 the likely location of restart markers, you may be able to do better.
1584 Read the read_restart_marker() and jpeg_resync_to_restart() routines
1585 in jdmarker.c if you think you'd like to implement your own resync
1588 term_source (j_decompress_ptr cinfo)
1589 Terminate source --- called by jpeg_finish_decompress() after all
1590 data has been read. Often a no-op.
1592 For both fill_input_buffer() and skip_input_data(), there is no such thing
1593 as an EOF return. If the end of the file has been reached, the routine has
1594 a choice of exiting via ERREXIT() or inserting fake data into the buffer.
1595 In most cases, generating a warning message and inserting a fake EOI marker
1596 is the best course of action --- this will allow the decompressor to output
1597 however much of the image is there. In pathological cases, the decompressor
1598 may swallow the EOI and again demand data ... just keep feeding it fake EOIs.
1599 jdatasrc.c illustrates the recommended error recovery behavior.
1601 term_source() is NOT called by jpeg_abort() or jpeg_destroy(). If you want
1602 the source manager to be cleaned up during an abort, you must do it yourself.
1604 You will also need code to create a jpeg_source_mgr struct, fill in its method
1605 pointers, and insert a pointer to the struct into the "src" field of the JPEG
1606 decompression object. This can be done in-line in your setup code if you
1607 like, but it's probably cleaner to provide a separate routine similar to the
1608 jpeg_stdio_src() or jpeg_mem_src() routines of the supplied source managers.
1610 For more information, consult the memory and stdio source and destination
1611 managers in jdatasrc.c and jdatadst.c.
1617 Some applications need to use the JPEG library as an incremental memory-to-
1618 memory filter: when the compressed data buffer is filled or emptied, they want
1619 control to return to the outer loop, rather than expecting that the buffer can
1620 be emptied or reloaded within the data source/destination manager subroutine.
1621 The library supports this need by providing an "I/O suspension" mode, which we
1622 describe in this section.
1624 The I/O suspension mode is not a panacea: nothing is guaranteed about the
1625 maximum amount of time spent in any one call to the library, so it will not
1626 eliminate response-time problems in single-threaded applications. If you
1627 need guaranteed response time, we suggest you "bite the bullet" and implement
1628 a real multi-tasking capability.
1630 To use I/O suspension, cooperation is needed between the calling application
1631 and the data source or destination manager; you will always need a custom
1632 source/destination manager. (Please read the previous section if you haven't
1633 already.) The basic idea is that the empty_output_buffer() or
1634 fill_input_buffer() routine is a no-op, merely returning FALSE to indicate
1635 that it has done nothing. Upon seeing this, the JPEG library suspends
1636 operation and returns to its caller. The surrounding application is
1637 responsible for emptying or refilling the work buffer before calling the
1640 Compression suspension:
1642 For compression suspension, use an empty_output_buffer() routine that returns
1643 FALSE; typically it will not do anything else. This will cause the
1644 compressor to return to the caller of jpeg_write_scanlines(), with the return
1645 value indicating that not all the supplied scanlines have been accepted.
1646 The application must make more room in the output buffer, adjust the output
1647 buffer pointer/count appropriately, and then call jpeg_write_scanlines()
1648 again, pointing to the first unconsumed scanline.
1650 When forced to suspend, the compressor will backtrack to a convenient stopping
1651 point (usually the start of the current MCU); it will regenerate some output
1652 data when restarted. Therefore, although empty_output_buffer() is only
1653 called when the buffer is filled, you should NOT write out the entire buffer
1654 after a suspension. Write only the data up to the current position of
1655 next_output_byte/free_in_buffer. The data beyond that point will be
1656 regenerated after resumption.
1658 Because of the backtracking behavior, a good-size output buffer is essential
1659 for efficiency; you don't want the compressor to suspend often. (In fact, an
1660 overly small buffer could lead to infinite looping, if a single MCU required
1661 more data than would fit in the buffer.) We recommend a buffer of at least
1662 several Kbytes. You may want to insert explicit code to ensure that you don't
1663 call jpeg_write_scanlines() unless there is a reasonable amount of space in
1664 the output buffer; in other words, flush the buffer before trying to compress
1667 The compressor does not allow suspension while it is trying to write JPEG
1668 markers at the beginning and end of the file. This means that:
1669 * At the beginning of a compression operation, there must be enough free
1670 space in the output buffer to hold the header markers (typically 600 or
1671 so bytes). The recommended buffer size is bigger than this anyway, so
1672 this is not a problem as long as you start with an empty buffer. However,
1673 this restriction might catch you if you insert large special markers, such
1674 as a JFIF thumbnail image, without flushing the buffer afterwards.
1675 * When you call jpeg_finish_compress(), there must be enough space in the
1676 output buffer to emit any buffered data and the final EOI marker. In the
1677 current implementation, half a dozen bytes should suffice for this, but
1678 for safety's sake we recommend ensuring that at least 100 bytes are free
1679 before calling jpeg_finish_compress().
1681 A more significant restriction is that jpeg_finish_compress() cannot suspend.
1682 This means you cannot use suspension with multi-pass operating modes, namely
1683 Huffman code optimization and multiple-scan output. Those modes write the
1684 whole file during jpeg_finish_compress(), which will certainly result in
1685 buffer overrun. (Note that this restriction applies only to compression,
1686 not decompression. The decompressor supports input suspension in all of its
1689 Decompression suspension:
1691 For decompression suspension, use a fill_input_buffer() routine that simply
1692 returns FALSE (except perhaps during error recovery, as discussed below).
1693 This will cause the decompressor to return to its caller with an indication
1694 that suspension has occurred. This can happen at four places:
1695 * jpeg_read_header(): will return JPEG_SUSPENDED.
1696 * jpeg_start_decompress(): will return FALSE, rather than its usual TRUE.
1697 * jpeg_read_scanlines(): will return the number of scanlines already
1698 completed (possibly 0).
1699 * jpeg_finish_decompress(): will return FALSE, rather than its usual TRUE.
1700 The surrounding application must recognize these cases, load more data into
1701 the input buffer, and repeat the call. In the case of jpeg_read_scanlines(),
1702 increment the passed pointers past any scanlines successfully read.
1704 Just as with compression, the decompressor will typically backtrack to a
1705 convenient restart point before suspending. When fill_input_buffer() is
1706 called, next_input_byte/bytes_in_buffer point to the current restart point,
1707 which is where the decompressor will backtrack to if FALSE is returned.
1708 The data beyond that position must NOT be discarded if you suspend; it needs
1709 to be re-read upon resumption. In most implementations, you'll need to shift
1710 this data down to the start of your work buffer and then load more data after
1711 it. Again, this behavior means that a several-Kbyte work buffer is essential
1712 for decent performance; furthermore, you should load a reasonable amount of
1713 new data before resuming decompression. (If you loaded, say, only one new
1714 byte each time around, you could waste a LOT of cycles.)
1716 The skip_input_data() source manager routine requires special care in a
1717 suspension scenario. This routine is NOT granted the ability to suspend the
1718 decompressor; it can decrement bytes_in_buffer to zero, but no more. If the
1719 requested skip distance exceeds the amount of data currently in the input
1720 buffer, then skip_input_data() must set bytes_in_buffer to zero and record the
1721 additional skip distance somewhere else. The decompressor will immediately
1722 call fill_input_buffer(), which should return FALSE, which will cause a
1723 suspension return. The surrounding application must then arrange to discard
1724 the recorded number of bytes before it resumes loading the input buffer.
1725 (Yes, this design is rather baroque, but it avoids complexity in the far more
1726 common case where a non-suspending source manager is used.)
1728 If the input data has been exhausted, we recommend that you emit a warning
1729 and insert dummy EOI markers just as a non-suspending data source manager
1730 would do. This can be handled either in the surrounding application logic or
1731 within fill_input_buffer(); the latter is probably more efficient. If
1732 fill_input_buffer() knows that no more data is available, it can set the
1733 pointer/count to point to a dummy EOI marker and then return TRUE just as
1734 though it had read more data in a non-suspending situation.
1736 The decompressor does not attempt to suspend within standard JPEG markers;
1737 instead it will backtrack to the start of the marker and reprocess the whole
1738 marker next time. Hence the input buffer must be large enough to hold the
1739 longest standard marker in the file. Standard JPEG markers should normally
1740 not exceed a few hundred bytes each (DHT tables are typically the longest).
1741 We recommend at least a 2K buffer for performance reasons, which is much
1742 larger than any correct marker is likely to be. For robustness against
1743 damaged marker length counts, you may wish to insert a test in your
1744 application for the case that the input buffer is completely full and yet
1745 the decoder has suspended without consuming any data --- otherwise, if this
1746 situation did occur, it would lead to an endless loop. (The library can't
1747 provide this test since it has no idea whether "the buffer is full", or
1748 even whether there is a fixed-size input buffer.)
1750 The input buffer would need to be 64K to allow for arbitrary COM or APPn
1751 markers, but these are handled specially: they are either saved into allocated
1752 memory, or skipped over by calling skip_input_data(). In the former case,
1753 suspension is handled correctly, and in the latter case, the problem of
1754 buffer overrun is placed on skip_input_data's shoulders, as explained above.
1755 Note that if you provide your own marker handling routine for large markers,
1756 you should consider how to deal with buffer overflow.
1758 Multiple-buffer management:
1760 In some applications it is desirable to store the compressed data in a linked
1761 list of buffer areas, so as to avoid data copying. This can be handled by
1762 having empty_output_buffer() or fill_input_buffer() set the pointer and count
1763 to reference the next available buffer; FALSE is returned only if no more
1764 buffers are available. Although seemingly straightforward, there is a
1765 pitfall in this approach: the backtrack that occurs when FALSE is returned
1766 could back up into an earlier buffer. For example, when fill_input_buffer()
1767 is called, the current pointer & count indicate the backtrack restart point.
1768 Since fill_input_buffer() will set the pointer and count to refer to a new
1769 buffer, the restart position must be saved somewhere else. Suppose a second
1770 call to fill_input_buffer() occurs in the same library call, and no
1771 additional input data is available, so fill_input_buffer must return FALSE.
1772 If the JPEG library has not moved the pointer/count forward in the current
1773 buffer, then *the correct restart point is the saved position in the prior
1774 buffer*. Prior buffers may be discarded only after the library establishes
1775 a restart point within a later buffer. Similar remarks apply for output into
1778 The library will never attempt to backtrack over a skip_input_data() call,
1779 so any skipped data can be permanently discarded. You still have to deal
1780 with the case of skipping not-yet-received data, however.
1782 It's much simpler to use only a single buffer; when fill_input_buffer() is
1783 called, move any unconsumed data (beyond the current pointer/count) down to
1784 the beginning of this buffer and then load new data into the remaining buffer
1785 space. This approach requires a little more data copying but is far easier
1789 Progressive JPEG support
1790 ------------------------
1792 Progressive JPEG rearranges the stored data into a series of scans of
1793 increasing quality. In situations where a JPEG file is transmitted across a
1794 slow communications link, a decoder can generate a low-quality image very
1795 quickly from the first scan, then gradually improve the displayed quality as
1796 more scans are received. The final image after all scans are complete is
1797 identical to that of a regular (sequential) JPEG file of the same quality
1798 setting. Progressive JPEG files are often slightly smaller than equivalent
1799 sequential JPEG files, but the possibility of incremental display is the main
1800 reason for using progressive JPEG.
1802 The IJG encoder library generates progressive JPEG files when given a
1803 suitable "scan script" defining how to divide the data into scans.
1804 Creation of progressive JPEG files is otherwise transparent to the encoder.
1805 Progressive JPEG files can also be read transparently by the decoder library.
1806 If the decoding application simply uses the library as defined above, it
1807 will receive a final decoded image without any indication that the file was
1808 progressive. Of course, this approach does not allow incremental display.
1809 To perform incremental display, an application needs to use the decoder
1810 library's "buffered-image" mode, in which it receives a decoded image
1813 Each displayed scan requires about as much work to decode as a full JPEG
1814 image of the same size, so the decoder must be fairly fast in relation to the
1815 data transmission rate in order to make incremental display useful. However,
1816 it is possible to skip displaying the image and simply add the incoming bits
1817 to the decoder's coefficient buffer. This is fast because only Huffman
1818 decoding need be done, not IDCT, upsampling, colorspace conversion, etc.
1819 The IJG decoder library allows the application to switch dynamically between
1820 displaying the image and simply absorbing the incoming bits. A properly
1821 coded application can automatically adapt the number of display passes to
1822 suit the time available as the image is received. Also, a final
1823 higher-quality display cycle can be performed from the buffered data after
1824 the end of the file is reached.
1826 Progressive compression:
1828 To create a progressive JPEG file (or a multiple-scan sequential JPEG file),
1829 set the scan_info cinfo field to point to an array of scan descriptors, and
1830 perform compression as usual. Instead of constructing your own scan list,
1831 you can call the jpeg_simple_progression() helper routine to create a
1832 recommended progression sequence; this method should be used by all
1833 applications that don't want to get involved in the nitty-gritty of
1834 progressive scan sequence design. (If you want to provide user control of
1835 scan sequences, you may wish to borrow the scan script reading code found
1836 in rdswitch.c, so that you can read scan script files just like cjpeg's.)
1837 When scan_info is not NULL, the compression library will store DCT'd data
1838 into a buffer array as jpeg_write_scanlines() is called, and will emit all
1839 the requested scans during jpeg_finish_compress(). This implies that
1840 multiple-scan output cannot be created with a suspending data destination
1841 manager, since jpeg_finish_compress() does not support suspension. We
1842 should also note that the compressor currently forces Huffman optimization
1843 mode when creating a progressive JPEG file, because the default Huffman
1844 tables are unsuitable for progressive files.
1846 Progressive decompression:
1848 When buffered-image mode is not used, the decoder library will read all of
1849 a multi-scan file during jpeg_start_decompress(), so that it can provide a
1850 final decoded image. (Here "multi-scan" means either progressive or
1851 multi-scan sequential.) This makes multi-scan files transparent to the
1852 decoding application. However, existing applications that used suspending
1853 input with version 5 of the IJG library will need to be modified to check
1854 for a suspension return from jpeg_start_decompress().
1856 To perform incremental display, an application must use the library's
1857 buffered-image mode. This is described in the next section.
1863 In buffered-image mode, the library stores the partially decoded image in a
1864 coefficient buffer, from which it can be read out as many times as desired.
1865 This mode is typically used for incremental display of progressive JPEG files,
1866 but it can be used with any JPEG file. Each scan of a progressive JPEG file
1867 adds more data (more detail) to the buffered image. The application can
1868 display in lockstep with the source file (one display pass per input scan),
1869 or it can allow input processing to outrun display processing. By making
1870 input and display processing run independently, it is possible for the
1871 application to adapt progressive display to a wide range of data transmission
1874 The basic control flow for buffered-image decoding is
1876 jpeg_create_decompress()
1879 set overall decompression parameters
1880 cinfo.buffered_image = TRUE; /* select buffered-image mode */
1881 jpeg_start_decompress()
1882 for (each output pass) {
1883 adjust output decompression parameters if required
1884 jpeg_start_output() /* start a new output pass */
1885 for (all scanlines in image) {
1886 jpeg_read_scanlines()
1889 jpeg_finish_output() /* terminate output pass */
1891 jpeg_finish_decompress()
1892 jpeg_destroy_decompress()
1894 This differs from ordinary unbuffered decoding in that there is an additional
1895 level of looping. The application can choose how many output passes to make
1896 and how to display each pass.
1898 The simplest approach to displaying progressive images is to do one display
1899 pass for each scan appearing in the input file. In this case the outer loop
1900 condition is typically
1901 while (! jpeg_input_complete(&cinfo))
1902 and the start-output call should read
1903 jpeg_start_output(&cinfo, cinfo.input_scan_number);
1904 The second parameter to jpeg_start_output() indicates which scan of the input
1905 file is to be displayed; the scans are numbered starting at 1 for this
1906 purpose. (You can use a loop counter starting at 1 if you like, but using
1907 the library's input scan counter is easier.) The library automatically reads
1908 data as necessary to complete each requested scan, and jpeg_finish_output()
1909 advances to the next scan or end-of-image marker (hence input_scan_number
1910 will be incremented by the time control arrives back at jpeg_start_output()).
1911 With this technique, data is read from the input file only as needed, and
1912 input and output processing run in lockstep.
1914 After reading the final scan and reaching the end of the input file, the
1915 buffered image remains available; it can be read additional times by
1916 repeating the jpeg_start_output()/jpeg_read_scanlines()/jpeg_finish_output()
1917 sequence. For example, a useful technique is to use fast one-pass color
1918 quantization for display passes made while the image is arriving, followed by
1919 a final display pass using two-pass quantization for highest quality. This
1920 is done by changing the library parameters before the final output pass.
1921 Changing parameters between passes is discussed in detail below.
1923 In general the last scan of a progressive file cannot be recognized as such
1924 until after it is read, so a post-input display pass is the best approach if
1925 you want special processing in the final pass.
1927 When done with the image, be sure to call jpeg_finish_decompress() to release
1928 the buffered image (or just use jpeg_destroy_decompress()).
1930 If input data arrives faster than it can be displayed, the application can
1931 cause the library to decode input data in advance of what's needed to produce
1932 output. This is done by calling the routine jpeg_consume_input().
1933 The return value is one of the following:
1934 JPEG_REACHED_SOS: reached an SOS marker (the start of a new scan)
1935 JPEG_REACHED_EOI: reached the EOI marker (end of image)
1936 JPEG_ROW_COMPLETED: completed reading one MCU row of compressed data
1937 JPEG_SCAN_COMPLETED: completed reading last MCU row of current scan
1938 JPEG_SUSPENDED: suspended before completing any of the above
1939 (JPEG_SUSPENDED can occur only if a suspending data source is used.) This
1940 routine can be called at any time after initializing the JPEG object. It
1941 reads some additional data and returns when one of the indicated significant
1942 events occurs. (If called after the EOI marker is reached, it will
1943 immediately return JPEG_REACHED_EOI without attempting to read more data.)
1945 The library's output processing will automatically call jpeg_consume_input()
1946 whenever the output processing overtakes the input; thus, simple lockstep
1947 display requires no direct calls to jpeg_consume_input(). But by adding
1948 calls to jpeg_consume_input(), you can absorb data in advance of what is
1949 being displayed. This has two benefits:
1950 * You can limit buildup of unprocessed data in your input buffer.
1951 * You can eliminate extra display passes by paying attention to the
1952 state of the library's input processing.
1954 The first of these benefits only requires interspersing calls to
1955 jpeg_consume_input() with your display operations and any other processing
1956 you may be doing. To avoid wasting cycles due to backtracking, it's best to
1957 call jpeg_consume_input() only after a hundred or so new bytes have arrived.
1958 This is discussed further under "I/O suspension", above. (Note: the JPEG
1959 library currently is not thread-safe. You must not call jpeg_consume_input()
1960 from one thread of control if a different library routine is working on the
1961 same JPEG object in another thread.)
1963 When input arrives fast enough that more than one new scan is available
1964 before you start a new output pass, you may as well skip the output pass
1965 corresponding to the completed scan. This occurs for free if you pass
1966 cinfo.input_scan_number as the target scan number to jpeg_start_output().
1967 The input_scan_number field is simply the index of the scan currently being
1968 consumed by the input processor. You can ensure that this is up-to-date by
1969 emptying the input buffer just before calling jpeg_start_output(): call
1970 jpeg_consume_input() repeatedly until it returns JPEG_SUSPENDED or
1973 The target scan number passed to jpeg_start_output() is saved in the
1974 cinfo.output_scan_number field. The library's output processing calls
1975 jpeg_consume_input() whenever the current input scan number and row within
1976 that scan is less than or equal to the current output scan number and row.
1977 Thus, input processing can "get ahead" of the output processing but is not
1978 allowed to "fall behind". You can achieve several different effects by
1979 manipulating this interlock rule. For example, if you pass a target scan
1980 number greater than the current input scan number, the output processor will
1981 wait until that scan starts to arrive before producing any output. (To avoid
1982 an infinite loop, the target scan number is automatically reset to the last
1983 scan number when the end of image is reached. Thus, if you specify a large
1984 target scan number, the library will just absorb the entire input file and
1985 then perform an output pass. This is effectively the same as what
1986 jpeg_start_decompress() does when you don't select buffered-image mode.)
1987 When you pass a target scan number equal to the current input scan number,
1988 the image is displayed no faster than the current input scan arrives. The
1989 final possibility is to pass a target scan number less than the current input
1990 scan number; this disables the input/output interlock and causes the output
1991 processor to simply display whatever it finds in the image buffer, without
1992 waiting for input. (However, the library will not accept a target scan
1993 number less than one, so you can't avoid waiting for the first scan.)
1995 When data is arriving faster than the output display processing can advance
1996 through the image, jpeg_consume_input() will store data into the buffered
1997 image beyond the point at which the output processing is reading data out
1998 again. If the input arrives fast enough, it may "wrap around" the buffer to
1999 the point where the input is more than one whole scan ahead of the output.
2000 If the output processing simply proceeds through its display pass without
2001 paying attention to the input, the effect seen on-screen is that the lower
2002 part of the image is one or more scans better in quality than the upper part.
2003 Then, when the next output scan is started, you have a choice of what target
2004 scan number to use. The recommended choice is to use the current input scan
2005 number at that time, which implies that you've skipped the output scans
2006 corresponding to the input scans that were completed while you processed the
2007 previous output scan. In this way, the decoder automatically adapts its
2008 speed to the arriving data, by skipping output scans as necessary to keep up
2009 with the arriving data.
2011 When using this strategy, you'll want to be sure that you perform a final
2012 output pass after receiving all the data; otherwise your last display may not
2013 be full quality across the whole screen. So the right outer loop logic is
2014 something like this:
2016 absorb any waiting input by calling jpeg_consume_input()
2017 final_pass = jpeg_input_complete(&cinfo);
2018 adjust output decompression parameters if required
2019 jpeg_start_output(&cinfo, cinfo.input_scan_number);
2021 jpeg_finish_output()
2022 } while (! final_pass);
2023 rather than quitting as soon as jpeg_input_complete() returns TRUE. This
2024 arrangement makes it simple to use higher-quality decoding parameters
2025 for the final pass. But if you don't want to use special parameters for
2026 the final pass, the right loop logic is like this:
2028 absorb any waiting input by calling jpeg_consume_input()
2029 jpeg_start_output(&cinfo, cinfo.input_scan_number);
2031 jpeg_finish_output()
2032 if (jpeg_input_complete(&cinfo) &&
2033 cinfo.input_scan_number == cinfo.output_scan_number)
2036 In this case you don't need to know in advance whether an output pass is to
2037 be the last one, so it's not necessary to have reached EOF before starting
2038 the final output pass; rather, what you want to test is whether the output
2039 pass was performed in sync with the final input scan. This form of the loop
2040 will avoid an extra output pass whenever the decoder is able (or nearly able)
2041 to keep up with the incoming data.
2043 When the data transmission speed is high, you might begin a display pass,
2044 then find that much or all of the file has arrived before you can complete
2045 the pass. (You can detect this by noting the JPEG_REACHED_EOI return code
2046 from jpeg_consume_input(), or equivalently by testing jpeg_input_complete().)
2047 In this situation you may wish to abort the current display pass and start a
2048 new one using the newly arrived information. To do so, just call
2049 jpeg_finish_output() and then start a new pass with jpeg_start_output().
2051 A variant strategy is to abort and restart display if more than one complete
2052 scan arrives during an output pass; this can be detected by noting
2053 JPEG_REACHED_SOS returns and/or examining cinfo.input_scan_number. This
2054 idea should be employed with caution, however, since the display process
2055 might never get to the bottom of the image before being aborted, resulting
2056 in the lower part of the screen being several passes worse than the upper.
2057 In most cases it's probably best to abort an output pass only if the whole
2058 file has arrived and you want to begin the final output pass immediately.
2060 When receiving data across a communication link, we recommend always using
2061 the current input scan number for the output target scan number; if a
2062 higher-quality final pass is to be done, it should be started (aborting any
2063 incomplete output pass) as soon as the end of file is received. However,
2064 many other strategies are possible. For example, the application can examine
2065 the parameters of the current input scan and decide whether to display it or
2066 not. If the scan contains only chroma data, one might choose not to use it
2067 as the target scan, expecting that the scan will be small and will arrive
2068 quickly. To skip to the next scan, call jpeg_consume_input() until it
2069 returns JPEG_REACHED_SOS or JPEG_REACHED_EOI. Or just use the next higher
2070 number as the target scan for jpeg_start_output(); but that method doesn't
2071 let you inspect the next scan's parameters before deciding to display it.
2074 In buffered-image mode, jpeg_start_decompress() never performs input and
2075 thus never suspends. An application that uses input suspension with
2076 buffered-image mode must be prepared for suspension returns from these
2078 * jpeg_start_output() performs input only if you request 2-pass quantization
2079 and the target scan isn't fully read yet. (This is discussed below.)
2080 * jpeg_read_scanlines(), as always, returns the number of scanlines that it
2081 was able to produce before suspending.
2082 * jpeg_finish_output() will read any markers following the target scan,
2083 up to the end of the file or the SOS marker that begins another scan.
2084 (But it reads no input if jpeg_consume_input() has already reached the
2085 end of the file or a SOS marker beyond the target output scan.)
2086 * jpeg_finish_decompress() will read until the end of file, and thus can
2087 suspend if the end hasn't already been reached (as can be tested by
2088 calling jpeg_input_complete()).
2089 jpeg_start_output(), jpeg_finish_output(), and jpeg_finish_decompress()
2090 all return TRUE if they completed their tasks, FALSE if they had to suspend.
2091 In the event of a FALSE return, the application must load more input data
2092 and repeat the call. Applications that use non-suspending data sources need
2093 not check the return values of these three routines.
2096 It is possible to change decoding parameters between output passes in the
2097 buffered-image mode. The decoder library currently supports only very
2098 limited changes of parameters. ONLY THE FOLLOWING parameter changes are
2099 allowed after jpeg_start_decompress() is called:
2100 * dct_method can be changed before each call to jpeg_start_output().
2101 For example, one could use a fast DCT method for early scans, changing
2102 to a higher quality method for the final scan.
2103 * dither_mode can be changed before each call to jpeg_start_output();
2104 of course this has no impact if not using color quantization. Typically
2105 one would use ordered dither for initial passes, then switch to
2106 Floyd-Steinberg dither for the final pass. Caution: changing dither mode
2107 can cause more memory to be allocated by the library. Although the amount
2108 of memory involved is not large (a scanline or so), it may cause the
2109 initial max_memory_to_use specification to be exceeded, which in the worst
2110 case would result in an out-of-memory failure.
2111 * do_block_smoothing can be changed before each call to jpeg_start_output().
2112 This setting is relevant only when decoding a progressive JPEG image.
2113 During the first DC-only scan, block smoothing provides a very "fuzzy" look
2114 instead of the very "blocky" look seen without it; which is better seems a
2115 matter of personal taste. But block smoothing is nearly always a win
2116 during later stages, especially when decoding a successive-approximation
2117 image: smoothing helps to hide the slight blockiness that otherwise shows
2118 up on smooth gradients until the lowest coefficient bits are sent.
2119 * Color quantization mode can be changed under the rules described below.
2120 You *cannot* change between full-color and quantized output (because that
2121 would alter the required I/O buffer sizes), but you can change which
2122 quantization method is used.
2124 When generating color-quantized output, changing quantization method is a
2125 very useful way of switching between high-speed and high-quality display.
2126 The library allows you to change among its three quantization methods:
2127 1. Single-pass quantization to a fixed color cube.
2128 Selected by cinfo.two_pass_quantize = FALSE and cinfo.colormap = NULL.
2129 2. Single-pass quantization to an application-supplied colormap.
2130 Selected by setting cinfo.colormap to point to the colormap (the value of
2131 two_pass_quantize is ignored); also set cinfo.actual_number_of_colors.
2132 3. Two-pass quantization to a colormap chosen specifically for the image.
2133 Selected by cinfo.two_pass_quantize = TRUE and cinfo.colormap = NULL.
2134 (This is the default setting selected by jpeg_read_header, but it is
2135 probably NOT what you want for the first pass of progressive display!)
2136 These methods offer successively better quality and lesser speed. However,
2137 only the first method is available for quantizing in non-RGB color spaces.
2139 IMPORTANT: because the different quantizer methods have very different
2140 working-storage requirements, the library requires you to indicate which
2141 one(s) you intend to use before you call jpeg_start_decompress(). (If we did
2142 not require this, the max_memory_to_use setting would be a complete fiction.)
2143 You do this by setting one or more of these three cinfo fields to TRUE:
2144 enable_1pass_quant Fixed color cube colormap
2145 enable_external_quant Externally-supplied colormap
2146 enable_2pass_quant Two-pass custom colormap
2147 All three are initialized FALSE by jpeg_read_header(). But
2148 jpeg_start_decompress() automatically sets TRUE the one selected by the
2149 current two_pass_quantize and colormap settings, so you only need to set the
2150 enable flags for any other quantization methods you plan to change to later.
2152 After setting the enable flags correctly at jpeg_start_decompress() time, you
2153 can change to any enabled quantization method by setting two_pass_quantize
2154 and colormap properly just before calling jpeg_start_output(). The following
2155 special rules apply:
2156 1. You must explicitly set cinfo.colormap to NULL when switching to 1-pass
2157 or 2-pass mode from a different mode, or when you want the 2-pass
2158 quantizer to be re-run to generate a new colormap.
2159 2. To switch to an external colormap, or to change to a different external
2160 colormap than was used on the prior pass, you must call
2161 jpeg_new_colormap() after setting cinfo.colormap.
2162 NOTE: if you want to use the same colormap as was used in the prior pass,
2163 you should not do either of these things. This will save some nontrivial
2165 (These requirements exist because cinfo.colormap will always be non-NULL
2166 after completing a prior output pass, since both the 1-pass and 2-pass
2167 quantizers set it to point to their output colormaps. Thus you have to
2168 do one of these two things to notify the library that something has changed.
2169 Yup, it's a bit klugy, but it's necessary to do it this way for backwards
2172 Note that in buffered-image mode, the library generates any requested colormap
2173 during jpeg_start_output(), not during jpeg_start_decompress().
2175 When using two-pass quantization, jpeg_start_output() makes a pass over the
2176 buffered image to determine the optimum color map; it therefore may take a
2177 significant amount of time, whereas ordinarily it does little work. The
2178 progress monitor hook is called during this pass, if defined. It is also
2179 important to realize that if the specified target scan number is greater than
2180 or equal to the current input scan number, jpeg_start_output() will attempt
2181 to consume input as it makes this pass. If you use a suspending data source,
2182 you need to check for a FALSE return from jpeg_start_output() under these
2183 conditions. The combination of 2-pass quantization and a not-yet-fully-read
2184 target scan is the only case in which jpeg_start_output() will consume input.
2187 Application authors who support buffered-image mode may be tempted to use it
2188 for all JPEG images, even single-scan ones. This will work, but it is
2189 inefficient: there is no need to create an image-sized coefficient buffer for
2190 single-scan images. Requesting buffered-image mode for such an image wastes
2191 memory. Worse, it can cost time on large images, since the buffered data has
2192 to be swapped out or written to a temporary file. If you are concerned about
2193 maximum performance on baseline JPEG files, you should use buffered-image
2194 mode only when the incoming file actually has multiple scans. This can be
2195 tested by calling jpeg_has_multiple_scans(), which will return a correct
2196 result at any time after jpeg_read_header() completes.
2198 It is also worth noting that when you use jpeg_consume_input() to let input
2199 processing get ahead of output processing, the resulting pattern of access to
2200 the coefficient buffer is quite nonsequential. It's best to use the memory
2201 manager jmemnobs.c if you can (ie, if you have enough real or virtual main
2202 memory). If not, at least make sure that max_memory_to_use is set as high as
2203 possible. If the JPEG memory manager has to use a temporary file, you will
2204 probably see a lot of disk traffic and poor performance. (This could be
2205 improved with additional work on the memory manager, but we haven't gotten
2208 In some applications it may be convenient to use jpeg_consume_input() for all
2209 input processing, including reading the initial markers; that is, you may
2210 wish to call jpeg_consume_input() instead of jpeg_read_header() during
2211 startup. This works, but note that you must check for JPEG_REACHED_SOS and
2212 JPEG_REACHED_EOI return codes as the equivalent of jpeg_read_header's codes.
2213 Once the first SOS marker has been reached, you must call
2214 jpeg_start_decompress() before jpeg_consume_input() will consume more input;
2215 it'll just keep returning JPEG_REACHED_SOS until you do. If you read a
2216 tables-only file this way, jpeg_consume_input() will return JPEG_REACHED_EOI
2217 without ever returning JPEG_REACHED_SOS; be sure to check for this case.
2218 If this happens, the decompressor will not read any more input until you call
2219 jpeg_abort() to reset it. It is OK to call jpeg_consume_input() even when not
2220 using buffered-image mode, but in that case it's basically a no-op after the
2221 initial markers have been read: it will just return JPEG_SUSPENDED.
2224 Abbreviated datastreams and multiple images
2225 -------------------------------------------
2227 A JPEG compression or decompression object can be reused to process multiple
2228 images. This saves a small amount of time per image by eliminating the
2229 "create" and "destroy" operations, but that isn't the real purpose of the
2230 feature. Rather, reuse of an object provides support for abbreviated JPEG
2231 datastreams. Object reuse can also simplify processing a series of images in
2232 a single input or output file. This section explains these features.
2234 A JPEG file normally contains several hundred bytes worth of quantization
2235 and Huffman tables. In a situation where many images will be stored or
2236 transmitted with identical tables, this may represent an annoying overhead.
2237 The JPEG standard therefore permits tables to be omitted. The standard
2238 defines three classes of JPEG datastreams:
2239 * "Interchange" datastreams contain an image and all tables needed to decode
2240 the image. These are the usual kind of JPEG file.
2241 * "Abbreviated image" datastreams contain an image, but are missing some or
2242 all of the tables needed to decode that image.
2243 * "Abbreviated table specification" (henceforth "tables-only") datastreams
2244 contain only table specifications.
2245 To decode an abbreviated image, it is necessary to load the missing table(s)
2246 into the decoder beforehand. This can be accomplished by reading a separate
2247 tables-only file. A variant scheme uses a series of images in which the first
2248 image is an interchange (complete) datastream, while subsequent ones are
2249 abbreviated and rely on the tables loaded by the first image. It is assumed
2250 that once the decoder has read a table, it will remember that table until a
2251 new definition for the same table number is encountered.
2253 It is the application designer's responsibility to figure out how to associate
2254 the correct tables with an abbreviated image. While abbreviated datastreams
2255 can be useful in a closed environment, their use is strongly discouraged in
2256 any situation where data exchange with other applications might be needed.
2259 The JPEG library provides support for reading and writing any combination of
2260 tables-only datastreams and abbreviated images. In both compression and
2261 decompression objects, a quantization or Huffman table will be retained for
2262 the lifetime of the object, unless it is overwritten by a new table definition.
2265 To create abbreviated image datastreams, it is only necessary to tell the
2266 compressor not to emit some or all of the tables it is using. Each
2267 quantization and Huffman table struct contains a boolean field "sent_table",
2268 which normally is initialized to FALSE. For each table used by the image, the
2269 header-writing process emits the table and sets sent_table = TRUE unless it is
2270 already TRUE. (In normal usage, this prevents outputting the same table
2271 definition multiple times, as would otherwise occur because the chroma
2272 components typically share tables.) Thus, setting this field to TRUE before
2273 calling jpeg_start_compress() will prevent the table from being written at
2276 If you want to create a "pure" abbreviated image file containing no tables,
2277 just call "jpeg_suppress_tables(&cinfo, TRUE)" after constructing all the
2278 tables. If you want to emit some but not all tables, you'll need to set the
2279 individual sent_table fields directly.
2281 To create an abbreviated image, you must also call jpeg_start_compress()
2282 with a second parameter of FALSE, not TRUE. Otherwise jpeg_start_compress()
2283 will force all the sent_table fields to FALSE. (This is a safety feature to
2284 prevent abbreviated images from being created accidentally.)
2286 To create a tables-only file, perform the same parameter setup that you
2287 normally would, but instead of calling jpeg_start_compress() and so on, call
2288 jpeg_write_tables(&cinfo). This will write an abbreviated datastream
2289 containing only SOI, DQT and/or DHT markers, and EOI. All the quantization
2290 and Huffman tables that are currently defined in the compression object will
2291 be emitted unless their sent_tables flag is already TRUE, and then all the
2292 sent_tables flags will be set TRUE.
2294 A sure-fire way to create matching tables-only and abbreviated image files
2295 is to proceed as follows:
2297 create JPEG compression object
2299 set destination to tables-only file
2300 jpeg_write_tables(&cinfo);
2301 set destination to image file
2302 jpeg_start_compress(&cinfo, FALSE);
2304 jpeg_finish_compress(&cinfo);
2306 Since the JPEG parameters are not altered between writing the table file and
2307 the abbreviated image file, the same tables are sure to be used. Of course,
2308 you can repeat the jpeg_start_compress() ... jpeg_finish_compress() sequence
2309 many times to produce many abbreviated image files matching the table file.
2311 You cannot suppress output of the computed Huffman tables when Huffman
2312 optimization is selected. (If you could, there'd be no way to decode the
2313 image...) Generally, you don't want to set optimize_coding = TRUE when
2314 you are trying to produce abbreviated files.
2316 In some cases you might want to compress an image using tables which are
2317 not stored in the application, but are defined in an interchange or
2318 tables-only file readable by the application. This can be done by setting up
2319 a JPEG decompression object to read the specification file, then copying the
2320 tables into your compression object. See jpeg_copy_critical_parameters()
2321 for an example of copying quantization tables.
2324 To read abbreviated image files, you simply need to load the proper tables
2325 into the decompression object before trying to read the abbreviated image.
2326 If the proper tables are stored in the application program, you can just
2327 allocate the table structs and fill in their contents directly. For example,
2328 to load a fixed quantization table into table slot "n":
2330 if (cinfo.quant_tbl_ptrs[n] == NULL)
2331 cinfo.quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) &cinfo);
2332 quant_ptr = cinfo.quant_tbl_ptrs[n]; /* quant_ptr is JQUANT_TBL* */
2333 for (i = 0; i < 64; i++) {
2334 /* Qtable[] is desired quantization table, in natural array order */
2335 quant_ptr->quantval[i] = Qtable[i];
2338 Code to load a fixed Huffman table is typically (for AC table "n"):
2340 if (cinfo.ac_huff_tbl_ptrs[n] == NULL)
2341 cinfo.ac_huff_tbl_ptrs[n] = jpeg_alloc_huff_table((j_common_ptr) &cinfo);
2342 huff_ptr = cinfo.ac_huff_tbl_ptrs[n]; /* huff_ptr is JHUFF_TBL* */
2343 for (i = 1; i <= 16; i++) {
2344 /* counts[i] is number of Huffman codes of length i bits, i=1..16 */
2345 huff_ptr->bits[i] = counts[i];
2347 for (i = 0; i < 256; i++) {
2348 /* symbols[] is the list of Huffman symbols, in code-length order */
2349 huff_ptr->huffval[i] = symbols[i];
2352 (Note that trying to set cinfo.quant_tbl_ptrs[n] to point directly at a
2353 constant JQUANT_TBL object is not safe. If the incoming file happened to
2354 contain a quantization table definition, your master table would get
2355 overwritten! Instead allocate a working table copy and copy the master table
2356 into it, as illustrated above. Ditto for Huffman tables, of course.)
2358 You might want to read the tables from a tables-only file, rather than
2359 hard-wiring them into your application. The jpeg_read_header() call is
2360 sufficient to read a tables-only file. You must pass a second parameter of
2361 FALSE to indicate that you do not require an image to be present. Thus, the
2364 create JPEG decompression object
2365 set source to tables-only file
2366 jpeg_read_header(&cinfo, FALSE);
2367 set source to abbreviated image file
2368 jpeg_read_header(&cinfo, TRUE);
2369 set decompression parameters
2370 jpeg_start_decompress(&cinfo);
2372 jpeg_finish_decompress(&cinfo);
2374 In some cases, you may want to read a file without knowing whether it contains
2375 an image or just tables. In that case, pass FALSE and check the return value
2376 from jpeg_read_header(): it will be JPEG_HEADER_OK if an image was found,
2377 JPEG_HEADER_TABLES_ONLY if only tables were found. (A third return value,
2378 JPEG_SUSPENDED, is possible when using a suspending data source manager.)
2379 Note that jpeg_read_header() will not complain if you read an abbreviated
2380 image for which you haven't loaded the missing tables; the missing-table check
2381 occurs later, in jpeg_start_decompress().
2384 It is possible to read a series of images from a single source file by
2385 repeating the jpeg_read_header() ... jpeg_finish_decompress() sequence,
2386 without releasing/recreating the JPEG object or the data source module.
2387 (If you did reinitialize, any partial bufferload left in the data source
2388 buffer at the end of one image would be discarded, causing you to lose the
2389 start of the next image.) When you use this method, stored tables are
2390 automatically carried forward, so some of the images can be abbreviated images
2391 that depend on tables from earlier images.
2393 If you intend to write a series of images into a single destination file,
2394 you might want to make a specialized data destination module that doesn't
2395 flush the output buffer at term_destination() time. This would speed things
2396 up by some trifling amount. Of course, you'd need to remember to flush the
2397 buffer after the last image. You can make the later images be abbreviated
2398 ones by passing FALSE to jpeg_start_compress().
2404 Some applications may need to insert or extract special data in the JPEG
2405 datastream. The JPEG standard provides marker types "COM" (comment) and
2406 "APP0" through "APP15" (application) to hold application-specific data.
2407 Unfortunately, the use of these markers is not specified by the standard.
2408 COM markers are fairly widely used to hold user-supplied text. The JFIF file
2409 format spec uses APP0 markers with specified initial strings to hold certain
2410 data. Adobe applications use APP14 markers beginning with the string "Adobe"
2411 for miscellaneous data. Other APPn markers are rarely seen, but might
2412 contain almost anything.
2414 If you wish to store user-supplied text, we recommend you use COM markers
2415 and place readable 7-bit ASCII text in them. Newline conventions are not
2416 standardized --- expect to find LF (Unix style), CR/LF (DOS style), or CR
2417 (Mac style). A robust COM reader should be able to cope with random binary
2418 garbage, including nulls, since some applications generate COM markers
2419 containing non-ASCII junk. (But yours should not be one of them.)
2421 For program-supplied data, use an APPn marker, and be sure to begin it with an
2422 identifying string so that you can tell whether the marker is actually yours.
2423 It's probably best to avoid using APP0 or APP14 for any private markers.
2424 (NOTE: the upcoming SPIFF standard will use APP8 markers; we recommend you
2425 not use APP8 markers for any private purposes, either.)
2427 Keep in mind that at most 65533 bytes can be put into one marker, but you
2428 can have as many markers as you like.
2430 By default, the IJG compression library will write a JFIF APP0 marker if the
2431 selected JPEG colorspace is grayscale or YCbCr, or an Adobe APP14 marker if
2432 the selected colorspace is RGB, CMYK, or YCCK. You can disable this, but
2433 we don't recommend it. The decompression library will recognize JFIF and
2434 Adobe markers and will set the JPEG colorspace properly when one is found.
2437 You can write special markers immediately following the datastream header by
2438 calling jpeg_write_marker() after jpeg_start_compress() and before the first
2439 call to jpeg_write_scanlines(). When you do this, the markers appear after
2440 the SOI and the JFIF APP0 and Adobe APP14 markers (if written), but before
2441 all else. Specify the marker type parameter as "JPEG_COM" for COM or
2442 "JPEG_APP0 + n" for APPn. (Actually, jpeg_write_marker will let you write
2443 any marker type, but we don't recommend writing any other kinds of marker.)
2444 For example, to write a user comment string pointed to by comment_text:
2445 jpeg_write_marker(cinfo, JPEG_COM, comment_text, strlen(comment_text));
2447 If it's not convenient to store all the marker data in memory at once,
2448 you can instead call jpeg_write_m_header() followed by multiple calls to
2449 jpeg_write_m_byte(). If you do it this way, it's your responsibility to
2450 call jpeg_write_m_byte() exactly the number of times given in the length
2451 parameter to jpeg_write_m_header(). (This method lets you empty the
2452 output buffer partway through a marker, which might be important when
2453 using a suspending data destination module. In any case, if you are using
2454 a suspending destination, you should flush its buffer after inserting
2455 any special markers. See "I/O suspension".)
2457 Or, if you prefer to synthesize the marker byte sequence yourself,
2458 you can just cram it straight into the data destination module.
2460 If you are writing JFIF 1.02 extension markers (thumbnail images), don't
2461 forget to set cinfo.JFIF_minor_version = 2 so that the encoder will write the
2462 correct JFIF version number in the JFIF header marker. The library's default
2463 is to write version 1.01, but that's wrong if you insert any 1.02 extension
2464 markers. (We could probably get away with just defaulting to 1.02, but there
2465 used to be broken decoders that would complain about unknown minor version
2466 numbers. To reduce compatibility risks it's safest not to write 1.02 unless
2467 you are actually using 1.02 extensions.)
2470 When reading, two methods of handling special markers are available:
2471 1. You can ask the library to save the contents of COM and/or APPn markers
2472 into memory, and then examine them at your leisure afterwards.
2473 2. You can supply your own routine to process COM and/or APPn markers
2474 on-the-fly as they are read.
2475 The first method is simpler to use, especially if you are using a suspending
2476 data source; writing a marker processor that copes with input suspension is
2477 not easy (consider what happens if the marker is longer than your available
2478 input buffer). However, the second method conserves memory since the marker
2479 data need not be kept around after it's been processed.
2481 For either method, you'd normally set up marker handling after creating a
2482 decompression object and before calling jpeg_read_header(), because the
2483 markers of interest will typically be near the head of the file and so will
2484 be scanned by jpeg_read_header. Once you've established a marker handling
2485 method, it will be used for the life of that decompression object
2486 (potentially many datastreams), unless you change it. Marker handling is
2487 determined separately for COM markers and for each APPn marker code.
2490 To save the contents of special markers in memory, call
2491 jpeg_save_markers(cinfo, marker_code, length_limit)
2492 where marker_code is the marker type to save, JPEG_COM or JPEG_APP0+n.
2493 (To arrange to save all the special marker types, you need to call this
2494 routine 17 times, for COM and APP0-APP15.) If the incoming marker is longer
2495 than length_limit data bytes, only length_limit bytes will be saved; this
2496 parameter allows you to avoid chewing up memory when you only need to see the
2497 first few bytes of a potentially large marker. If you want to save all the
2498 data, set length_limit to 0xFFFF; that is enough since marker lengths are only
2499 16 bits. As a special case, setting length_limit to 0 prevents that marker
2500 type from being saved at all. (That is the default behavior, in fact.)
2502 After jpeg_read_header() completes, you can examine the special markers by
2503 following the cinfo->marker_list pointer chain. All the special markers in
2504 the file appear in this list, in order of their occurrence in the file (but
2505 omitting any markers of types you didn't ask for). Both the original data
2506 length and the saved data length are recorded for each list entry; the latter
2507 will not exceed length_limit for the particular marker type. Note that these
2508 lengths exclude the marker length word, whereas the stored representation
2509 within the JPEG file includes it. (Hence the maximum data length is really
2512 It is possible that additional special markers appear in the file beyond the
2513 SOS marker at which jpeg_read_header stops; if so, the marker list will be
2514 extended during reading of the rest of the file. This is not expected to be
2515 common, however. If you are short on memory you may want to reset the length
2516 limit to zero for all marker types after finishing jpeg_read_header, to
2517 ensure that the max_memory_to_use setting cannot be exceeded due to addition
2520 The marker list remains stored until you call jpeg_finish_decompress or
2521 jpeg_abort, at which point the memory is freed and the list is set to empty.
2522 (jpeg_destroy also releases the storage, of course.)
2524 Note that the library is internally interested in APP0 and APP14 markers;
2525 if you try to set a small nonzero length limit on these types, the library
2526 will silently force the length up to the minimum it wants. (But you can set
2527 a zero length limit to prevent them from being saved at all.) Also, in a
2528 16-bit environment, the maximum length limit may be constrained to less than
2529 65533 by malloc() limitations. It is therefore best not to assume that the
2530 effective length limit is exactly what you set it to be.
2533 If you want to supply your own marker-reading routine, you do it by calling
2534 jpeg_set_marker_processor(). A marker processor routine must have the
2536 boolean jpeg_marker_parser_method (j_decompress_ptr cinfo)
2537 Although the marker code is not explicitly passed, the routine can find it
2538 in cinfo->unread_marker. At the time of call, the marker proper has been
2539 read from the data source module. The processor routine is responsible for
2540 reading the marker length word and the remaining parameter bytes, if any.
2541 Return TRUE to indicate success. (FALSE should be returned only if you are
2542 using a suspending data source and it tells you to suspend. See the standard
2543 marker processors in jdmarker.c for appropriate coding methods if you need to
2544 use a suspending data source.)
2546 If you override the default APP0 or APP14 processors, it is up to you to
2547 recognize JFIF and Adobe markers if you want colorspace recognition to occur
2548 properly. We recommend copying and extending the default processors if you
2549 want to do that. (A better idea is to save these marker types for later
2550 examination by calling jpeg_save_markers(); that method doesn't interfere
2551 with the library's own processing of these markers.)
2553 jpeg_set_marker_processor() and jpeg_save_markers() are mutually exclusive
2554 --- if you call one it overrides any previous call to the other, for the
2555 particular marker type specified.
2557 A simple example of an external COM processor can be found in djpeg.c.
2558 Also, see jpegtran.c for an example of using jpeg_save_markers.
2561 Raw (downsampled) image data
2562 ----------------------------
2564 Some applications need to supply already-downsampled image data to the JPEG
2565 compressor, or to receive raw downsampled data from the decompressor. The
2566 library supports this requirement by allowing the application to write or
2567 read raw data, bypassing the normal preprocessing or postprocessing steps.
2568 The interface is different from the standard one and is somewhat harder to
2569 use. If your interest is merely in bypassing color conversion, we recommend
2570 that you use the standard interface and simply set jpeg_color_space =
2571 in_color_space (or jpeg_color_space = out_color_space for decompression).
2572 The mechanism described in this section is necessary only to supply or
2573 receive downsampled image data, in which not all components have the same
2577 To compress raw data, you must supply the data in the colorspace to be used
2578 in the JPEG file (please read the earlier section on Special color spaces)
2579 and downsampled to the sampling factors specified in the JPEG parameters.
2580 You must supply the data in the format used internally by the JPEG library,
2581 namely a JSAMPIMAGE array. This is an array of pointers to two-dimensional
2582 arrays, each of type JSAMPARRAY. Each 2-D array holds the values for one
2583 color component. This structure is necessary since the components are of
2584 different sizes. If the image dimensions are not a multiple of the MCU size,
2585 you must also pad the data correctly (usually, this is done by replicating
2586 the last column and/or row). The data must be padded to a multiple of a DCT
2587 block in each component: that is, each downsampled row must contain a
2588 multiple of 8 valid samples, and there must be a multiple of 8 sample rows
2589 for each component. (For applications such as conversion of digital TV
2590 images, the standard image size is usually a multiple of the DCT block size,
2591 so that no padding need actually be done.)
2593 The procedure for compression of raw data is basically the same as normal
2594 compression, except that you call jpeg_write_raw_data() in place of
2595 jpeg_write_scanlines(). Before calling jpeg_start_compress(), you must do
2597 * Set cinfo->raw_data_in to TRUE. (It is set FALSE by jpeg_set_defaults().)
2598 This notifies the library that you will be supplying raw data.
2599 Furthermore, set cinfo->do_fancy_downsampling to FALSE if you want to use
2600 real downsampled data. (It is set TRUE by jpeg_set_defaults().)
2601 * Ensure jpeg_color_space is correct --- an explicit jpeg_set_colorspace()
2602 call is a good idea. Note that since color conversion is bypassed,
2603 in_color_space is ignored, except that jpeg_set_defaults() uses it to
2604 choose the default jpeg_color_space setting.
2605 * Ensure the sampling factors, cinfo->comp_info[i].h_samp_factor and
2606 cinfo->comp_info[i].v_samp_factor, are correct. Since these indicate the
2607 dimensions of the data you are supplying, it's wise to set them
2608 explicitly, rather than assuming the library's defaults are what you want.
2610 To pass raw data to the library, call jpeg_write_raw_data() in place of
2611 jpeg_write_scanlines(). The two routines work similarly except that
2612 jpeg_write_raw_data takes a JSAMPIMAGE data array rather than JSAMPARRAY.
2613 The scanlines count passed to and returned from jpeg_write_raw_data is
2614 measured in terms of the component with the largest v_samp_factor.
2616 jpeg_write_raw_data() processes one MCU row per call, which is to say
2617 v_samp_factor*DCTSIZE sample rows of each component. The passed num_lines
2618 value must be at least max_v_samp_factor*DCTSIZE, and the return value will
2619 be exactly that amount (or possibly some multiple of that amount, in future
2620 library versions). This is true even on the last call at the bottom of the
2621 image; don't forget to pad your data as necessary.
2623 The required dimensions of the supplied data can be computed for each
2625 cinfo->comp_info[i].width_in_blocks*DCTSIZE samples per row
2626 cinfo->comp_info[i].height_in_blocks*DCTSIZE rows in image
2627 after jpeg_start_compress() has initialized those fields. If the valid data
2628 is smaller than this, it must be padded appropriately. For some sampling
2629 factors and image sizes, additional dummy DCT blocks are inserted to make
2630 the image a multiple of the MCU dimensions. The library creates such dummy
2631 blocks itself; it does not read them from your supplied data. Therefore you
2632 need never pad by more than DCTSIZE samples. An example may help here.
2633 Assume 2h2v downsampling of YCbCr data, that is
2634 cinfo->comp_info[0].h_samp_factor = 2 for Y
2635 cinfo->comp_info[0].v_samp_factor = 2
2636 cinfo->comp_info[1].h_samp_factor = 1 for Cb
2637 cinfo->comp_info[1].v_samp_factor = 1
2638 cinfo->comp_info[2].h_samp_factor = 1 for Cr
2639 cinfo->comp_info[2].v_samp_factor = 1
2640 and suppose that the nominal image dimensions (cinfo->image_width and
2641 cinfo->image_height) are 101x101 pixels. Then jpeg_start_compress() will
2642 compute downsampled_width = 101 and width_in_blocks = 13 for Y,
2643 downsampled_width = 51 and width_in_blocks = 7 for Cb and Cr (and the same
2644 for the height fields). You must pad the Y data to at least 13*8 = 104
2645 columns and rows, the Cb/Cr data to at least 7*8 = 56 columns and rows. The
2646 MCU height is max_v_samp_factor = 2 DCT rows so you must pass at least 16
2647 scanlines on each call to jpeg_write_raw_data(), which is to say 16 actual
2648 sample rows of Y and 8 each of Cb and Cr. A total of 7 MCU rows are needed,
2649 so you must pass a total of 7*16 = 112 "scanlines". The last DCT block row
2650 of Y data is dummy, so it doesn't matter what you pass for it in the data
2651 arrays, but the scanlines count must total up to 112 so that all of the Cb
2652 and Cr data gets passed.
2654 Output suspension is supported with raw-data compression: if the data
2655 destination module suspends, jpeg_write_raw_data() will return 0.
2656 In this case the same data rows must be passed again on the next call.
2659 Decompression with raw data output implies bypassing all postprocessing.
2660 You must deal with the color space and sampling factors present in the
2661 incoming file. If your application only handles, say, 2h1v YCbCr data,
2662 you must check for and fail on other color spaces or other sampling factors.
2663 The library will not convert to a different color space for you.
2665 To obtain raw data output, set cinfo->raw_data_out = TRUE before
2666 jpeg_start_decompress() (it is set FALSE by jpeg_read_header()). Be sure to
2667 verify that the color space and sampling factors are ones you can handle.
2668 Furthermore, set cinfo->do_fancy_upsampling = FALSE if you want to get real
2669 downsampled data (it is set TRUE by jpeg_read_header()).
2670 Then call jpeg_read_raw_data() in place of jpeg_read_scanlines(). The
2671 decompression process is otherwise the same as usual.
2673 jpeg_read_raw_data() returns one MCU row per call, and thus you must pass a
2674 buffer of at least max_v_samp_factor*DCTSIZE scanlines (scanline counting is
2675 the same as for raw-data compression). The buffer you pass must be large
2676 enough to hold the actual data plus padding to DCT-block boundaries. As with
2677 compression, any entirely dummy DCT blocks are not processed so you need not
2678 allocate space for them, but the total scanline count includes them. The
2679 above example of computing buffer dimensions for raw-data compression is
2680 equally valid for decompression.
2682 Input suspension is supported with raw-data decompression: if the data source
2683 module suspends, jpeg_read_raw_data() will return 0. You can also use
2684 buffered-image mode to read raw data in multiple passes.
2687 Really raw data: DCT coefficients
2688 ---------------------------------
2690 It is possible to read or write the contents of a JPEG file as raw DCT
2691 coefficients. This facility is mainly intended for use in lossless
2692 transcoding between different JPEG file formats. Other possible applications
2693 include lossless cropping of a JPEG image, lossless reassembly of a
2694 multi-strip or multi-tile TIFF/JPEG file into a single JPEG datastream, etc.
2696 To read the contents of a JPEG file as DCT coefficients, open the file and do
2697 jpeg_read_header() as usual. But instead of calling jpeg_start_decompress()
2698 and jpeg_read_scanlines(), call jpeg_read_coefficients(). This will read the
2699 entire image into a set of virtual coefficient-block arrays, one array per
2700 component. The return value is a pointer to an array of virtual-array
2701 descriptors. Each virtual array can be accessed directly using the JPEG
2702 memory manager's access_virt_barray method (see Memory management, below,
2703 and also read structure.txt's discussion of virtual array handling). Or,
2704 for simple transcoding to a different JPEG file format, the array list can
2705 just be handed directly to jpeg_write_coefficients().
2707 Each block in the block arrays contains quantized coefficient values in
2708 normal array order (not JPEG zigzag order). The block arrays contain only
2709 DCT blocks containing real data; any entirely-dummy blocks added to fill out
2710 interleaved MCUs at the right or bottom edges of the image are discarded
2711 during reading and are not stored in the block arrays. (The size of each
2712 block array can be determined from the width_in_blocks and height_in_blocks
2713 fields of the component's comp_info entry.) This is also the data format
2714 expected by jpeg_write_coefficients().
2716 When you are done using the virtual arrays, call jpeg_finish_decompress()
2717 to release the array storage and return the decompression object to an idle
2718 state; or just call jpeg_destroy() if you don't need to reuse the object.
2720 If you use a suspending data source, jpeg_read_coefficients() will return
2721 NULL if it is forced to suspend; a non-NULL return value indicates successful
2722 completion. You need not test for a NULL return value when using a
2723 non-suspending data source.
2725 It is also possible to call jpeg_read_coefficients() to obtain access to the
2726 decoder's coefficient arrays during a normal decode cycle in buffered-image
2727 mode. This frammish might be useful for progressively displaying an incoming
2728 image and then re-encoding it without loss. To do this, decode in buffered-
2729 image mode as discussed previously, then call jpeg_read_coefficients() after
2730 the last jpeg_finish_output() call. The arrays will be available for your use
2731 until you call jpeg_finish_decompress().
2734 To write the contents of a JPEG file as DCT coefficients, you must provide
2735 the DCT coefficients stored in virtual block arrays. You can either pass
2736 block arrays read from an input JPEG file by jpeg_read_coefficients(), or
2737 allocate virtual arrays from the JPEG compression object and fill them
2738 yourself. In either case, jpeg_write_coefficients() is substituted for
2739 jpeg_start_compress() and jpeg_write_scanlines(). Thus the sequence is
2740 * Create compression object
2741 * Set all compression parameters as necessary
2742 * Request virtual arrays if needed
2743 * jpeg_write_coefficients()
2744 * jpeg_finish_compress()
2745 * Destroy or re-use compression object
2746 jpeg_write_coefficients() is passed a pointer to an array of virtual block
2747 array descriptors; the number of arrays is equal to cinfo.num_components.
2749 The virtual arrays need only have been requested, not realized, before
2750 jpeg_write_coefficients() is called. A side-effect of
2751 jpeg_write_coefficients() is to realize any virtual arrays that have been
2752 requested from the compression object's memory manager. Thus, when obtaining
2753 the virtual arrays from the compression object, you should fill the arrays
2754 after calling jpeg_write_coefficients(). The data is actually written out
2755 when you call jpeg_finish_compress(); jpeg_write_coefficients() only writes
2758 When writing raw DCT coefficients, it is crucial that the JPEG quantization
2759 tables and sampling factors match the way the data was encoded, or the
2760 resulting file will be invalid. For transcoding from an existing JPEG file,
2761 we recommend using jpeg_copy_critical_parameters(). This routine initializes
2762 all the compression parameters to default values (like jpeg_set_defaults()),
2763 then copies the critical information from a source decompression object.
2764 The decompression object should have just been used to read the entire
2765 JPEG input file --- that is, it should be awaiting jpeg_finish_decompress().
2767 jpeg_write_coefficients() marks all tables stored in the compression object
2768 as needing to be written to the output file (thus, it acts like
2769 jpeg_start_compress(cinfo, TRUE)). This is for safety's sake, to avoid
2770 emitting abbreviated JPEG files by accident. If you really want to emit an
2771 abbreviated JPEG file, call jpeg_suppress_tables(), or set the tables'
2772 individual sent_table flags, between calling jpeg_write_coefficients() and
2773 jpeg_finish_compress().
2779 Some applications may need to regain control from the JPEG library every so
2780 often. The typical use of this feature is to produce a percent-done bar or
2781 other progress display. (For a simple example, see cjpeg.c or djpeg.c.)
2782 Although you do get control back frequently during the data-transferring pass
2783 (the jpeg_read_scanlines or jpeg_write_scanlines loop), any additional passes
2784 will occur inside jpeg_finish_compress or jpeg_start_decompress; those
2785 routines may take a long time to execute, and you don't get control back
2786 until they are done.
2788 You can define a progress-monitor routine which will be called periodically
2789 by the library. No guarantees are made about how often this call will occur,
2790 so we don't recommend you use it for mouse tracking or anything like that.
2791 At present, a call will occur once per MCU row, scanline, or sample row
2792 group, whichever unit is convenient for the current processing mode; so the
2793 wider the image, the longer the time between calls. During the data
2794 transferring pass, only one call occurs per call of jpeg_read_scanlines or
2795 jpeg_write_scanlines, so don't pass a large number of scanlines at once if
2796 you want fine resolution in the progress count. (If you really need to use
2797 the callback mechanism for time-critical tasks like mouse tracking, you could
2798 insert additional calls inside some of the library's inner loops.)
2800 To establish a progress-monitor callback, create a struct jpeg_progress_mgr,
2801 fill in its progress_monitor field with a pointer to your callback routine,
2802 and set cinfo->progress to point to the struct. The callback will be called
2803 whenever cinfo->progress is non-NULL. (This pointer is set to NULL by
2804 jpeg_create_compress or jpeg_create_decompress; the library will not change
2805 it thereafter. So if you allocate dynamic storage for the progress struct,
2806 make sure it will live as long as the JPEG object does. Allocating from the
2807 JPEG memory manager with lifetime JPOOL_PERMANENT will work nicely.) You
2808 can use the same callback routine for both compression and decompression.
2810 The jpeg_progress_mgr struct contains four fields which are set by the library:
2811 long pass_counter; /* work units completed in this pass */
2812 long pass_limit; /* total number of work units in this pass */
2813 int completed_passes; /* passes completed so far */
2814 int total_passes; /* total number of passes expected */
2815 During any one pass, pass_counter increases from 0 up to (not including)
2816 pass_limit; the step size is usually but not necessarily 1. The pass_limit
2817 value may change from one pass to another. The expected total number of
2818 passes is in total_passes, and the number of passes already completed is in
2819 completed_passes. Thus the fraction of work completed may be estimated as
2820 completed_passes + (pass_counter/pass_limit)
2821 --------------------------------------------
2823 ignoring the fact that the passes may not be equal amounts of work.
2825 When decompressing, pass_limit can even change within a pass, because it
2826 depends on the number of scans in the JPEG file, which isn't always known in
2827 advance. The computed fraction-of-work-done may jump suddenly (if the library
2828 discovers it has overestimated the number of scans) or even decrease (in the
2829 opposite case). It is not wise to put great faith in the work estimate.
2831 When using the decompressor's buffered-image mode, the progress monitor work
2832 estimate is likely to be completely unhelpful, because the library has no way
2833 to know how many output passes will be demanded of it. Currently, the library
2834 sets total_passes based on the assumption that there will be one more output
2835 pass if the input file end hasn't yet been read (jpeg_input_complete() isn't
2836 TRUE), but no more output passes if the file end has been reached when the
2837 output pass is started. This means that total_passes will rise as additional
2838 output passes are requested. If you have a way of determining the input file
2839 size, estimating progress based on the fraction of the file that's been read
2840 will probably be more useful than using the library's value.
2846 This section covers some key facts about the JPEG library's built-in memory
2847 manager. For more info, please read structure.txt's section about the memory
2848 manager, and consult the source code if necessary.
2850 All memory and temporary file allocation within the library is done via the
2851 memory manager. If necessary, you can replace the "back end" of the memory
2852 manager to control allocation yourself (for example, if you don't want the
2853 library to use malloc() and free() for some reason).
2855 Some data is allocated "permanently" and will not be freed until the JPEG
2856 object is destroyed. Most data is allocated "per image" and is freed by
2857 jpeg_finish_compress, jpeg_finish_decompress, or jpeg_abort. You can call the
2858 memory manager yourself to allocate structures that will automatically be
2859 freed at these times. Typical code for this is
2860 ptr = (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, size);
2861 Use JPOOL_PERMANENT to get storage that lasts as long as the JPEG object.
2862 Use alloc_large instead of alloc_small for anything bigger than a few Kbytes.
2863 There are also alloc_sarray and alloc_barray routines that automatically
2864 build 2-D sample or block arrays.
2866 The library's minimum space requirements to process an image depend on the
2867 image's width, but not on its height, because the library ordinarily works
2868 with "strip" buffers that are as wide as the image but just a few rows high.
2869 Some operating modes (eg, two-pass color quantization) require full-image
2870 buffers. Such buffers are treated as "virtual arrays": only the current strip
2871 need be in memory, and the rest can be swapped out to a temporary file.
2873 If you use the simplest memory manager back end (jmemnobs.c), then no
2874 temporary files are used; virtual arrays are simply malloc()'d. Images bigger
2875 than memory can be processed only if your system supports virtual memory.
2876 The other memory manager back ends support temporary files of various flavors
2877 and thus work in machines without virtual memory. They may also be useful on
2878 Unix machines if you need to process images that exceed available swap space.
2880 When using temporary files, the library will make the in-memory buffers for
2881 its virtual arrays just big enough to stay within a "maximum memory" setting.
2882 Your application can set this limit by setting cinfo->mem->max_memory_to_use
2883 after creating the JPEG object. (Of course, there is still a minimum size for
2884 the buffers, so the max-memory setting is effective only if it is bigger than
2885 the minimum space needed.) If you allocate any large structures yourself, you
2886 must allocate them before jpeg_start_compress() or jpeg_start_decompress() in
2887 order to have them counted against the max memory limit. Also keep in mind
2888 that space allocated with alloc_small() is ignored, on the assumption that
2889 it's too small to be worth worrying about; so a reasonable safety margin
2890 should be left when setting max_memory_to_use.
2892 If you use the jmemname.c or jmemdos.c memory manager back end, it is
2893 important to clean up the JPEG object properly to ensure that the temporary
2894 files get deleted. (This is especially crucial with jmemdos.c, where the
2895 "temporary files" may be extended-memory segments; if they are not freed,
2896 DOS will require a reboot to recover the memory.) Thus, with these memory
2897 managers, it's a good idea to provide a signal handler that will trap any
2898 early exit from your program. The handler should call either jpeg_abort()
2899 or jpeg_destroy() for any active JPEG objects. A handler is not needed with
2900 jmemnobs.c, and shouldn't be necessary with jmemansi.c or jmemmac.c either,
2901 since the C library is supposed to take care of deleting files made with
2908 Working memory requirements while performing compression or decompression
2909 depend on image dimensions, image characteristics (such as colorspace and
2910 JPEG process), and operating mode (application-selected options).
2912 As of v6b, the decompressor requires:
2913 1. About 24K in more-or-less-fixed-size data. This varies a bit depending
2914 on operating mode and image characteristics (particularly color vs.
2915 grayscale), but it doesn't depend on image dimensions.
2916 2. Strip buffers (of size proportional to the image width) for IDCT and
2917 upsampling results. The worst case for commonly used sampling factors
2918 is about 34 bytes * width in pixels for a color image. A grayscale image
2919 only needs about 8 bytes per pixel column.
2920 3. A full-image DCT coefficient buffer is needed to decode a multi-scan JPEG
2921 file (including progressive JPEGs), or whenever you select buffered-image
2922 mode. This takes 2 bytes/coefficient. At typical 2x2 sampling, that's
2923 3 bytes per pixel for a color image. Worst case (1x1 sampling) requires
2924 6 bytes/pixel. For grayscale, figure 2 bytes/pixel.
2925 4. To perform 2-pass color quantization, the decompressor also needs a
2926 128K color lookup table and a full-image pixel buffer (3 bytes/pixel).
2927 This does not count any memory allocated by the application, such as a
2928 buffer to hold the final output image.
2930 The above figures are valid for 8-bit JPEG data precision and a machine with
2931 32-bit ints. For 12-bit JPEG data, double the size of the strip buffers and
2932 quantization pixel buffer. The "fixed-size" data will be somewhat smaller
2933 with 16-bit ints, larger with 64-bit ints. Also, CMYK or other unusual
2934 color spaces will require different amounts of space.
2936 The full-image coefficient and pixel buffers, if needed at all, do not
2937 have to be fully RAM resident; you can have the library use temporary
2938 files instead when the total memory usage would exceed a limit you set.
2939 (But if your OS supports virtual memory, it's probably better to just use
2940 jmemnobs and let the OS do the swapping.)
2942 The compressor's memory requirements are similar, except that it has no need
2943 for color quantization. Also, it needs a full-image DCT coefficient buffer
2944 if Huffman-table optimization is asked for, even if progressive mode is not
2947 If you need more detailed information about memory usage in a particular
2948 situation, you can enable the MEM_STATS code in jmemmgr.c.
2951 Library compile-time options
2952 ----------------------------
2954 A number of compile-time options are available by modifying jmorecfg.h.
2956 The JPEG standard provides for both the baseline 8-bit DCT process and
2957 a 12-bit DCT process. The IJG code supports 12-bit JPEG if you define
2958 BITS_IN_JSAMPLE as 12 rather than 8. Note that this causes JSAMPLE to be
2959 larger than a char, so it affects the surrounding application's image data.
2960 The sample applications cjpeg and djpeg can support 12-bit mode only for PPM
2961 and GIF file formats; you must disable the other file formats to compile a
2962 12-bit cjpeg or djpeg. (install.txt has more information about that.)
2963 At present, a 12-bit library can handle *only* 12-bit images, not both
2964 precisions. (If you need to include both 8- and 12-bit libraries in a single
2965 application, you could probably do it by defining NEED_SHORT_EXTERNAL_NAMES
2966 for just one of the copies. You'd have to access the 8-bit and 12-bit copies
2967 from separate application source files. This is untested ... if you try it,
2968 we'd like to hear whether it works!)
2970 Note that a 12-bit library always compresses in Huffman optimization mode,
2971 in order to generate valid Huffman tables. This is necessary because our
2972 default Huffman tables only cover 8-bit data. If you need to output 12-bit
2973 files in one pass, you'll have to supply suitable default Huffman tables.
2974 You may also want to supply your own DCT quantization tables; the existing
2975 quality-scaling code has been developed for 8-bit use, and probably doesn't
2976 generate especially good tables for 12-bit.
2978 The maximum number of components (color channels) in the image is determined
2979 by MAX_COMPONENTS. The JPEG standard allows up to 255 components, but we
2980 expect that few applications will need more than four or so.
2982 On machines with unusual data type sizes, you may be able to improve
2983 performance or reduce memory space by tweaking the various typedefs in
2984 jmorecfg.h. In particular, on some RISC CPUs, access to arrays of "short"s
2985 is quite slow; consider trading memory for speed by making JCOEF, INT16, and
2986 UINT16 be "int" or "unsigned int". UINT8 is also a candidate to become int.
2987 You probably don't want to make JSAMPLE be int unless you have lots of memory
2990 You can reduce the size of the library by compiling out various optional
2991 functions. To do this, undefine xxx_SUPPORTED symbols as necessary.
2993 You can also save a few K by not having text error messages in the library;
2994 the standard error message table occupies about 5Kb. This is particularly
2995 reasonable for embedded applications where there's no good way to display
2996 a message anyway. To do this, remove the creation of the message table
2997 (jpeg_std_message_table[]) from jerror.c, and alter format_message to do
2998 something reasonable without it. You could output the numeric value of the
2999 message code number, for example. If you do this, you can also save a couple
3000 more K by modifying the TRACEMSn() macros in jerror.h to expand to nothing;
3001 you don't need trace capability anyway, right?
3004 Portability considerations
3005 --------------------------
3007 The JPEG library has been written to be extremely portable; the sample
3008 applications cjpeg and djpeg are slightly less so. This section summarizes
3009 the design goals in this area. (If you encounter any bugs that cause the
3010 library to be less portable than is claimed here, we'd appreciate hearing
3013 The code works fine on ANSI C, C++, and pre-ANSI C compilers, using any of
3014 the popular system include file setups, and some not-so-popular ones too.
3015 See install.txt for configuration procedures.
3017 The code is not dependent on the exact sizes of the C data types. As
3018 distributed, we make the assumptions that
3019 char is at least 8 bits wide
3020 short is at least 16 bits wide
3021 int is at least 16 bits wide
3022 long is at least 32 bits wide
3023 (These are the minimum requirements of the ANSI C standard.) Wider types will
3024 work fine, although memory may be used inefficiently if char is much larger
3025 than 8 bits or short is much bigger than 16 bits. The code should work
3026 equally well with 16- or 32-bit ints.
3028 In a system where these assumptions are not met, you may be able to make the
3029 code work by modifying the typedefs in jmorecfg.h. However, you will probably
3030 have difficulty if int is less than 16 bits wide, since references to plain
3031 int abound in the code.
3033 char can be either signed or unsigned, although the code runs faster if an
3034 unsigned char type is available. If char is wider than 8 bits, you will need
3035 to redefine JOCTET and/or provide custom data source/destination managers so
3036 that JOCTET represents exactly 8 bits of data on external storage.
3038 The JPEG library proper does not assume ASCII representation of characters.
3039 But some of the image file I/O modules in cjpeg/djpeg do have ASCII
3040 dependencies in file-header manipulation; so does cjpeg's select_file_type()
3043 The JPEG library does not rely heavily on the C library. In particular, C
3044 stdio is used only by the data source/destination modules and the error
3045 handler, all of which are application-replaceable. (cjpeg/djpeg are more
3046 heavily dependent on stdio.) malloc and free are called only from the memory
3047 manager "back end" module, so you can use a different memory allocator by
3048 replacing that one file.
3050 The code generally assumes that C names must be unique in the first 15
3051 characters. However, global function names can be made unique in the
3052 first 6 characters by defining NEED_SHORT_EXTERNAL_NAMES.
3054 More info about porting the code may be gleaned by reading jconfig.txt,
3055 jmorecfg.h, and jinclude.h.
3058 Notes for MS-DOS implementors
3059 -----------------------------
3061 The IJG code is designed to work efficiently in 80x86 "small" or "medium"
3062 memory models (i.e., data pointers are 16 bits unless explicitly declared
3063 "far"; code pointers can be either size). You may be able to use small
3064 model to compile cjpeg or djpeg by itself, but you will probably have to use
3065 medium model for any larger application. This won't make much difference in
3066 performance. You *will* take a noticeable performance hit if you use a
3067 large-data memory model (perhaps 10%-25%), and you should avoid "huge" model
3070 The JPEG library typically needs 2Kb-3Kb of stack space. It will also
3071 malloc about 20K-30K of near heap space while executing (and lots of far
3072 heap, but that doesn't count in this calculation). This figure will vary
3073 depending on selected operating mode, and to a lesser extent on image size.
3074 There is also about 5Kb-6Kb of constant data which will be allocated in the
3075 near data segment (about 4Kb of this is the error message table).
3076 Thus you have perhaps 20K available for other modules' static data and near
3077 heap space before you need to go to a larger memory model. The C library's
3078 static data will account for several K of this, but that still leaves a good
3079 deal for your needs. (If you are tight on space, you could reduce the sizes
3080 of the I/O buffers allocated by jdatasrc.c and jdatadst.c, say from 4K to
3081 1K. Another possibility is to move the error message table to far memory;
3082 this should be doable with only localized hacking on jerror.c.)
3084 About 2K of the near heap space is "permanent" memory that will not be
3085 released until you destroy the JPEG object. This is only an issue if you
3086 save a JPEG object between compression or decompression operations.
3088 Far data space may also be a tight resource when you are dealing with large
3089 images. The most memory-intensive case is decompression with two-pass color
3090 quantization, or single-pass quantization to an externally supplied color
3091 map. This requires a 128Kb color lookup table plus strip buffers amounting
3092 to about 40 bytes per column for typical sampling ratios (eg, about 25600
3093 bytes for a 640-pixel-wide image). You may not be able to process wide
3094 images if you have large data structures of your own.
3096 Of course, all of these concerns vanish if you use a 32-bit flat-memory-model
3097 compiler, such as DJGPP or Watcom C. We highly recommend flat model if you
3098 can use it; the JPEG library is significantly faster in flat model.