media/base/h264_bitstream_converter.cc

   1 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "media/base/h264_bitstream_converter.h"
   6
   7 #include "base/logging.h"
   8
   9 namespace media {
  10
  11 static const uint8 kStartCodePrefix[3] = {0, 0, 1};
  12
  13 // Helper function which determines whether NAL unit of given type marks
  14 // access unit boundary.
  15 static bool IsAccessUnitBoundaryNal(int nal_unit_type) {
  16   // Check if this packet marks access unit boundary by checking the
  17   // packet type.
  18   if (nal_unit_type == 6 ||  // Supplemental enhancement information
  19       nal_unit_type == 7 ||  // Picture parameter set
  20       nal_unit_type == 8 ||  // Sequence parameter set
  21       nal_unit_type == 9 ||  // Access unit delimiter
  22       (nal_unit_type >= 14 && nal_unit_type <= 18)) {  // Reserved types
  23     return true;
  24   }
  25   return false;
  26 }
  27
  28 H264BitstreamConverter::H264BitstreamConverter()
  29     : configuration_processed_(false),
  30       first_nal_unit_in_access_unit_(true),
  31       nal_unit_length_field_width_(0) {
  32 }
  33
  34 H264BitstreamConverter::~H264BitstreamConverter() {}
  35
  36 uint32 H264BitstreamConverter::ParseConfigurationAndCalculateSize(
  37     const uint8* configuration_record,
  38     uint32 configuration_record_size) {
  39   // FFmpeg's AVCodecContext's extradata field contains the Decoder Specific
  40   // Information from MP4 headers that contain the H.264 SPS and PPS members.
  41   // ISO 14496-15 Chapter 5.2.4 AVCDecoderConfigurationRecord.
  42   // AVCConfigurationRecord must be at least 7 bytes long.
  43   if (configuration_record == NULL || configuration_record_size < 7) {
  44     return 0;  // Error: invalid input
  45   }
  46   const uint8* decoder_configuration = configuration_record;
  47   uint32 parameter_set_size_bytes = 0;
  48
  49   // We can skip the four first bytes as they're only profile information
  50   decoder_configuration += 4;
  51   // Fifth byte's two LSBs contain the interleaving field's size minus one
  52   uint8 size_of_len_field = (*decoder_configuration & 0x3) + 1;
  53   if (size_of_len_field != 1 && size_of_len_field != 2 &&
  54       size_of_len_field != 4) {
  55     return 0;  // Error: invalid input, NAL unit field len is not correct
  56   }
  57   decoder_configuration++;
  58   // Sixth byte's five LSBs contain the number of SPSs
  59   uint8 sps_count = *decoder_configuration & 0x1F;
  60   decoder_configuration++;
  61   // Then we have N * SPS's with two byte length field and actual SPS
  62   while (sps_count-- > 0) {
  63     if ((decoder_configuration - configuration_record) + 2 >
  64          static_cast<int32>(configuration_record_size)) {
  65       return 0;  // Error: ran out of data
  66     }
  67     uint16 sps_len = decoder_configuration[0] << 8 | decoder_configuration[1];
  68     decoder_configuration += 2;
  69     // write the SPS to output, always with zero byte + start code prefix
  70     parameter_set_size_bytes += 1 + sizeof(kStartCodePrefix);
  71     decoder_configuration += sps_len;
  72     parameter_set_size_bytes += sps_len;
  73   }
  74   // Then we have the numner of pps in one byte
  75   uint8 pps_count = *decoder_configuration;
  76   decoder_configuration++;
  77   // And finally, we have N * PPS with two byte length field and actual PPS
  78   while (pps_count-- > 0) {
  79     if ((decoder_configuration - configuration_record) + 2 >
  80          static_cast<int32>(configuration_record_size)) {
  81       return 0;  // Error: ran out of data
  82     }
  83     uint16 pps_len = decoder_configuration[0] << 8 | decoder_configuration[1];
  84     decoder_configuration += 2;
  85     // write the SPS to output, always with zero byte + start code prefix
  86     parameter_set_size_bytes += 1 + sizeof(kStartCodePrefix);
  87     decoder_configuration += pps_len;
  88     parameter_set_size_bytes += pps_len;
  89   }
  90   // We're done processing the AVCDecoderConfigurationRecord,
  91   // store the needed information for parsing actual payload
  92   nal_unit_length_field_width_ = size_of_len_field;
  93   configuration_processed_ = true;
  94   return parameter_set_size_bytes;
  95 }
  96
  97 uint32 H264BitstreamConverter::CalculateNeededOutputBufferSize(
  98     const uint8* input,
  99     uint32 input_size) const {
 100   uint32 output_size = 0;
 101   uint32 data_left = input_size;
 102   bool first_nal_in_this_access_unit = first_nal_unit_in_access_unit_;
 103
 104   if (input == NULL || input_size == 0) {
 105     return 0;  // Error: invalid input data
 106   }
 107   if (!configuration_processed_) {
 108     return 0;  // Error: configuration not handled, we don't know nal unit width
 109   }
 110   CHECK(nal_unit_length_field_width_ == 1 ||
 111         nal_unit_length_field_width_ == 2 ||
 112         nal_unit_length_field_width_ == 4);
 113
 114   // Then add the needed size for the actual packet
 115   while (data_left > 0) {
 116     // Read the next NAL unit length from the input buffer
 117     uint8 size_of_len_field;
 118     uint32 nal_unit_length;
 119     for (nal_unit_length = 0, size_of_len_field = nal_unit_length_field_width_;
 120          size_of_len_field > 0;
 121          input++, size_of_len_field--, data_left--) {
 122       nal_unit_length <<= 8;
 123       nal_unit_length |= *input;
 124     }
 125
 126     if (nal_unit_length == 0) {
 127       break;  // Signifies that no more data left in the buffer
 128     } else if (nal_unit_length > data_left) {
 129       return 0;  // Error: Not enough data for correct conversion
 130     }
 131     data_left -= nal_unit_length;
 132
 133     // five least significant bits of first NAL unit byte signify nal_unit_type
 134     int nal_unit_type = *input & 0x1F;
 135     if (first_nal_in_this_access_unit ||
 136         IsAccessUnitBoundaryNal(nal_unit_type)) {
 137       output_size += 1;  // Extra zero_byte for these nal units
 138       first_nal_in_this_access_unit = false;
 139     }
 140     // Start code prefix
 141     output_size += sizeof(kStartCodePrefix);
 142     // Actual NAL unit size
 143     output_size += nal_unit_length;
 144     input += nal_unit_length;
 145     // No need for trailing zero bits
 146   }
 147   return output_size;
 148 }
 149
 150 bool H264BitstreamConverter::ConvertAVCDecoderConfigurationRecordToByteStream(
 151     const uint8* input,
 152     uint32 input_size,
 153     uint8* output,
 154     uint32* output_size) {
 155   uint8* outscan = output;
 156   // FFmpeg's AVCodecContext's extradata field contains the Decoder Specific
 157   // Information from MP4 headers that contain the H.264 SPS and PPS members.
 158   // ISO 14496-15 Chapter 5.2.4 AVCDecoderConfigurationRecord.
 159   const uint8* decoder_configuration = input;
 160   uint32 decoderconfiguration_size = input_size;
 161   uint32 out_size = 0;
 162
 163   if (decoder_configuration == NULL || decoderconfiguration_size == 0) {
 164     return 0;  // Error: input invalid
 165   }
 166
 167   // We can skip the four first bytes as they're only profile information.
 168   decoder_configuration += 4;
 169   // Fifth byte's two LSBs contain the interleaving field's size minus one
 170   uint8 size_of_len_field = (*decoder_configuration & 0x3) + 1;
 171   if (size_of_len_field != 1 && size_of_len_field != 2 &&
 172       size_of_len_field != 4) {
 173     return 0;  // Error: invalid input, NAL unit field len is not correct
 174   }
 175   decoder_configuration++;
 176   // Sixth byte's five LSBs contain the number of SPSs
 177   uint8 sps_count = *decoder_configuration & 0x1F;
 178   decoder_configuration++;
 179   // Then we have N * SPS's with two byte length field and actual SPS
 180   while (sps_count-- > 0) {
 181     uint16 sps_len = decoder_configuration[0] << 8 |
 182                      decoder_configuration[1];
 183     decoder_configuration += 2;
 184     if (out_size + 1 + sizeof(kStartCodePrefix) + sps_len >
 185         *output_size) {
 186       *output_size = 0;
 187       return 0;  // too small output buffer;
 188     }
 189     // write the SPS to output, always with zero byte + start code prefix
 190     *outscan = 0;  // zero byte
 191     outscan += 1;
 192     memcpy(outscan, kStartCodePrefix, sizeof(kStartCodePrefix));
 193     outscan += sizeof(kStartCodePrefix);
 194     memcpy(outscan, decoder_configuration, sps_len);
 195     decoder_configuration += sps_len;
 196     outscan += sps_len;
 197     out_size += 1 + sizeof(kStartCodePrefix) + sps_len;
 198   }
 199   // Then we have the numner of pps in one byte
 200   uint8 pps_count = *decoder_configuration;
 201   decoder_configuration++;
 202   // And finally, we have N * PPS with two byte length field and actual PPS
 203   while (pps_count-- > 0) {
 204     uint16 pps_len = decoder_configuration[0] << 8 | decoder_configuration[1];
 205     decoder_configuration += 2;
 206     if (out_size + 1 + sizeof(kStartCodePrefix) + pps_len >
 207         *output_size) {
 208       *output_size = 0;
 209       return 0;  // too small output buffer;
 210     }
 211     // write the SPS to output, always with zero byte + start code prefix
 212     *outscan = 0;  // zero byte
 213     outscan += 1;
 214     memcpy(outscan, kStartCodePrefix, sizeof(kStartCodePrefix));
 215     outscan += sizeof(kStartCodePrefix);
 216     memcpy(outscan, decoder_configuration, pps_len);
 217     decoder_configuration += pps_len;
 218     outscan += pps_len;
 219     out_size += 1 + sizeof(kStartCodePrefix) + pps_len;
 220   }
 221   // We're done processing the AVCDecoderConfigurationRecord, store the needed
 222   // information
 223   nal_unit_length_field_width_ = size_of_len_field;
 224   configuration_processed_ = true;
 225   *output_size = out_size;
 226   return true;
 227 }
 228
 229 bool H264BitstreamConverter::ConvertNalUnitStreamToByteStream(
 230     const uint8* input, uint32 input_size,
 231     uint8* output, uint32* output_size) {
 232   const uint8* inscan = input;  // We read the input from here progressively
 233   uint8* outscan = output;  // We write the output to here progressively
 234   uint32 data_left = input_size;
 235
 236   if (inscan == NULL || input_size == 0 ||
 237       outscan == NULL || *output_size == 0) {
 238     *output_size = 0;
 239     return false;  // Error: invalid input
 240   }
 241
 242   // NAL unit width should be known at this point
 243   CHECK(nal_unit_length_field_width_ == 1 ||
 244         nal_unit_length_field_width_ == 2 ||
 245         nal_unit_length_field_width_ == 4);
 246
 247   // Do the actual conversion for the actual input packet
 248   while (data_left > 0) {
 249     uint8 i;
 250     uint32 nal_unit_length;
 251
 252     // Read the next NAL unit length from the input buffer by scanning
 253     // the input stream with the specific length field width
 254     for (nal_unit_length = 0, i = nal_unit_length_field_width_;
 255          i > 0 && data_left > 0;
 256          inscan++, i--, data_left--) {
 257       nal_unit_length <<= 8;
 258       nal_unit_length |= *inscan;
 259     }
 260
 261     if (nal_unit_length == 0) {
 262       break;  // Successful conversion, end of buffer
 263     } else if (nal_unit_length > data_left) {
 264       *output_size = 0;
 265       return false;  // Error: not enough data for correct conversion
 266     }
 267
 268     uint32 start_code_len;
 269     first_nal_unit_in_access_unit_ ?
 270         start_code_len = sizeof(kStartCodePrefix) + 1 :
 271         start_code_len = sizeof(kStartCodePrefix);
 272     if (static_cast<uint32>(outscan - output) +
 273         start_code_len + nal_unit_length > *output_size) {
 274       *output_size = 0;
 275       return false;  // Error: too small output buffer
 276     }
 277
 278     // Five least significant bits of first NAL unit byte signify
 279     // nal_unit_type.
 280     int nal_unit_type = *inscan & 0x1F;
 281
 282     // Check if this packet marks access unit boundary by checking the
 283     // packet type.
 284     if (IsAccessUnitBoundaryNal(nal_unit_type)) {
 285       first_nal_unit_in_access_unit_ = true;
 286     }
 287
 288     // Write extra zero-byte before start code prefix if this packet
 289     // signals next access unit.
 290     if (first_nal_unit_in_access_unit_) {
 291       *outscan = 0;
 292       outscan++;
 293       first_nal_unit_in_access_unit_ = false;
 294     }
 295
 296     // No need to write leading zero bits.
 297     // Write start-code prefix.
 298     memcpy(outscan, kStartCodePrefix, sizeof(kStartCodePrefix));
 299     outscan += sizeof(kStartCodePrefix);
 300     // Then write the actual NAL unit from the input buffer.
 301     memcpy(outscan, inscan, nal_unit_length);
 302     inscan += nal_unit_length;
 303     data_left -= nal_unit_length;
 304     outscan += nal_unit_length;
 305     // No need for trailing zero bits.
 306   }
 307   // Successful conversion, output the freshly allocated bitstream buffer.
 308   *output_size = static_cast<uint32>(outscan - output);
 309   return true;
 310 }
 311
 312 }  // namespace media