r123: Merged HEAD and TEST. New stuff shall be committed to HEAD from now on.
[cinelerra_cv/mob.git] / libmpeg3 / video / idct.c
blobab1f2badfba1937c279e45c27d0c295a8471f6a3
1 #include "idct.h"
2 #include <stdlib.h>
4 /**********************************************************/
5 /* inverse two dimensional DCT, Chen-Wang algorithm */
6 /* (cf. IEEE ASSP-32, pp. 803-816, Aug. 1984) */
7 /* 32-bit integer arithmetic (8 bit coefficients) */
8 /* 11 mults, 29 adds per DCT */
9 /* sE, 18.8.91 */
10 /**********************************************************/
11 /* coefficients extended to 12 bit for IEEE1180-1990 */
12 /* compliance sE, 2.1.94 */
13 /**********************************************************/
15 /* this code assumes >> to be a two's-complement arithmetic */
16 /* right shift: (-2)>>1 == -1 , (-3)>>1 == -2 */
18 #define W1 2841 /* 2048*sqrt(2)*cos(1*pi/16) */
19 #define W2 2676 /* 2048*sqrt(2)*cos(2*pi/16) */
20 #define W3 2408 /* 2048*sqrt(2)*cos(3*pi/16) */
21 #define W5 1609 /* 2048*sqrt(2)*cos(5*pi/16) */
22 #define W6 1108 /* 2048*sqrt(2)*cos(6*pi/16) */
23 #define W7 565 /* 2048*sqrt(2)*cos(7*pi/16) */
25 /* row (horizontal) IDCT
27 * 7 pi 1
28 * dst[k] = sum c[l] * src[l] * cos( -- * ( k + - ) * l )
29 * l=0 8 2
31 * where: c[0] = 128
32 * c[1..7] = 128*sqrt(2)
35 int mpeg3video_idctrow(short *blk)
37 int x0, x1, x2, x3, x4, x5, x6, x7, x8;
39 /* shortcut */
40 if (!((x1 = blk[4]<<11) | (x2 = blk[6]) | (x3 = blk[2]) |
41 (x4 = blk[1]) | (x5 = blk[7]) | (x6 = blk[5]) | (x7 = blk[3])))
43 blk[0]=blk[1]=blk[2]=blk[3]=blk[4]=blk[5]=blk[6]=blk[7]=blk[0]<<3;
44 return 0;
47 x0 = (blk[0]<<11) + 128; /* for proper rounding in the fourth stage */
49 /* first stage */
50 x8 = W7*(x4+x5);
51 x4 = x8 + (W1-W7)*x4;
52 x5 = x8 - (W1+W7)*x5;
53 x8 = W3*(x6+x7);
54 x6 = x8 - (W3-W5)*x6;
55 x7 = x8 - (W3+W5)*x7;
57 /* second stage */
58 x8 = x0 + x1;
59 x0 -= x1;
60 x1 = W6*(x3+x2);
61 x2 = x1 - (W2+W6)*x2;
62 x3 = x1 + (W2-W6)*x3;
63 x1 = x4 + x6;
64 x4 -= x6;
65 x6 = x5 + x7;
66 x5 -= x7;
68 /* third stage */
69 x7 = x8 + x3;
70 x8 -= x3;
71 x3 = x0 + x2;
72 x0 -= x2;
73 x2 = (181*(x4+x5)+128)>>8;
74 x4 = (181*(x4-x5)+128)>>8;
76 /* fourth stage */
77 blk[0] = (x7+x1)>>8;
78 blk[1] = (x3+x2)>>8;
79 blk[2] = (x0+x4)>>8;
80 blk[3] = (x8+x6)>>8;
81 blk[4] = (x8-x6)>>8;
82 blk[5] = (x0-x4)>>8;
83 blk[6] = (x3-x2)>>8;
84 blk[7] = (x7-x1)>>8;
85 return 0;
88 /* column (vertical) IDCT
90 * 7 pi 1
91 * dst[8*k] = sum c[l] * src[8*l] * cos( -- * ( k + - ) * l )
92 * l=0 8 2
94 * where: c[0] = 1/1024
95 * c[1..7] = (1/1024)*sqrt(2)
98 int mpeg3video_idctcol(short *blk)
100 int x0, x1, x2, x3, x4, x5, x6, x7, x8;
102 /* shortcut */
103 if (!((x1 = (blk[8 * 4]<<8)) | (x2 = blk[8 * 6]) | (x3 = blk[8 * 2]) |
104 (x4 = blk[8*1]) | (x5 = blk[8 * 7]) | (x6 = blk[8 * 5]) | (x7 = blk[8 * 3]))){
105 blk[8*0]=blk[8*1]=blk[8 * 2]=blk[8 * 3]=blk[8 * 4]=blk[8 * 5]=blk[8 * 6]=blk[8 * 7]=
106 (blk[8*0]+32)>>6;
107 return 0;
110 x0 = (blk[8*0]<<8) + 8192;
112 /* first stage */
113 x8 = W7*(x4+x5) + 4;
114 x4 = (x8+(W1-W7)*x4)>>3;
115 x5 = (x8-(W1+W7)*x5)>>3;
116 x8 = W3*(x6+x7) + 4;
117 x6 = (x8-(W3-W5)*x6)>>3;
118 x7 = (x8-(W3+W5)*x7)>>3;
120 /* second stage */
121 x8 = x0 + x1;
122 x0 -= x1;
123 x1 = W6*(x3+x2) + 4;
124 x2 = (x1-(W2+W6)*x2)>>3;
125 x3 = (x1+(W2-W6)*x3)>>3;
126 x1 = x4 + x6;
127 x4 -= x6;
128 x6 = x5 + x7;
129 x5 -= x7;
131 /* third stage */
132 x7 = x8 + x3;
133 x8 -= x3;
134 x3 = x0 + x2;
135 x0 -= x2;
136 x2 = (181 * (x4 + x5) + 128) >> 8;
137 x4 = (181 * (x4 - x5) + 128) >> 8;
139 /* fourth stage */
140 blk[8 * 0] = (x7 + x1) >> 14;
141 blk[8 * 1] = (x3 + x2) >> 14;
142 blk[8 * 2] = (x0 + x4) >> 14;
143 blk[8 * 3] = (x8 + x6) >> 14;
144 blk[8 * 4] = (x8 - x6) >> 14;
145 blk[8 * 5] = (x0 - x4) >> 14;
146 blk[8 * 6] = (x3 - x2) >> 14;
147 blk[8 * 7] = (x7 - x1) >> 14;
148 return 0;
152 /* two dimensional inverse discrete cosine transform */
153 void mpeg3video_idct_conversion(short* block)
155 int i;
156 for(i = 0; i < 8; i++) mpeg3video_idctrow(block + 8 * i);
157 for(i = 0; i < 8; i++) mpeg3video_idctcol(block + i);