media/formats/mp2t/es_parser_h264.cc

   1 // Copyright 2014 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/formats/mp2t/es_parser_h264.h"
   6
   7 #include "base/logging.h"
   8 #include "base/numerics/safe_conversions.h"
   9 #include "media/base/buffers.h"
  10 #include "media/base/stream_parser_buffer.h"
  11 #include "media/base/video_frame.h"
  12 #include "media/filters/h264_parser.h"
  13 #include "media/formats/common/offset_byte_queue.h"
  14 #include "media/formats/mp2t/mp2t_common.h"
  15 #include "ui/gfx/geometry/rect.h"
  16 #include "ui/gfx/geometry/size.h"
  17
  18 namespace media {
  19 namespace mp2t {
  20
  21 // An AUD NALU is at least 4 bytes:
  22 // 3 bytes for the start code + 1 byte for the NALU type.
  23 const int kMinAUDSize = 4;
  24
  25 EsParserH264::EsParserH264(
  26     const NewVideoConfigCB& new_video_config_cb,
  27     const EmitBufferCB& emit_buffer_cb)
  28     : es_adapter_(new_video_config_cb, emit_buffer_cb),
  29       h264_parser_(new H264Parser()),
  30       current_access_unit_pos_(0),
  31       next_access_unit_pos_(0) {
  32 }
  33
  34 EsParserH264::~EsParserH264() {
  35 }
  36
  37 void EsParserH264::Flush() {
  38   DVLOG(1) << __FUNCTION__;
  39   if (!FindAUD(&current_access_unit_pos_))
  40     return;
  41
  42   // Simulate an additional AUD to force emitting the last access unit
  43   // which is assumed to be complete at this point.
  44   uint8 aud[] = { 0x00, 0x00, 0x01, 0x09 };
  45   es_queue_->Push(aud, sizeof(aud));
  46   ParseFromEsQueue();
  47
  48   es_adapter_.Flush();
  49 }
  50
  51 void EsParserH264::ResetInternal() {
  52   DVLOG(1) << __FUNCTION__;
  53   h264_parser_.reset(new H264Parser());
  54   current_access_unit_pos_ = 0;
  55   next_access_unit_pos_ = 0;
  56   last_video_decoder_config_ = VideoDecoderConfig();
  57   es_adapter_.Reset();
  58 }
  59
  60 bool EsParserH264::FindAUD(int64* stream_pos) {
  61   while (true) {
  62     const uint8* es;
  63     int size;
  64     es_queue_->PeekAt(*stream_pos, &es, &size);
  65
  66     // Find a start code and move the stream to the start code parser position.
  67     off_t start_code_offset;
  68     off_t start_code_size;
  69     bool start_code_found = H264Parser::FindStartCode(
  70         es, size, &start_code_offset, &start_code_size);
  71     *stream_pos += start_code_offset;
  72
  73     // No H264 start code found or NALU type not available yet.
  74     if (!start_code_found || start_code_offset + start_code_size >= size)
  75       return false;
  76
  77     // Exit the parser loop when an AUD is found.
  78     // Note: NALU header for an AUD:
  79     // - nal_ref_idc must be 0
  80     // - nal_unit_type must be H264NALU::kAUD
  81     if (es[start_code_offset + start_code_size] == H264NALU::kAUD)
  82       break;
  83
  84     // The current NALU is not an AUD, skip the start code
  85     // and continue parsing the stream.
  86     *stream_pos += start_code_size;
  87   }
  88
  89   return true;
  90 }
  91
  92 bool EsParserH264::ParseFromEsQueue() {
  93   DCHECK_LE(es_queue_->head(), current_access_unit_pos_);
  94   DCHECK_LE(current_access_unit_pos_, next_access_unit_pos_);
  95   DCHECK_LE(next_access_unit_pos_, es_queue_->tail());
  96
  97   // Find the next AUD located at or after |current_access_unit_pos_|. This is
  98   // needed since initially |current_access_unit_pos_| might not point to
  99   // an AUD.
 100   // Discard all the data before the updated |current_access_unit_pos_|
 101   // since it won't be used again.
 102   bool aud_found = FindAUD(&current_access_unit_pos_);
 103   es_queue_->Trim(current_access_unit_pos_);
 104   if (next_access_unit_pos_ < current_access_unit_pos_)
 105     next_access_unit_pos_ = current_access_unit_pos_;
 106
 107   // Resume parsing later if no AUD was found.
 108   if (!aud_found)
 109     return true;
 110
 111   // Find the next AUD to make sure we have a complete access unit.
 112   if (next_access_unit_pos_ < current_access_unit_pos_ + kMinAUDSize) {
 113     next_access_unit_pos_ = current_access_unit_pos_ + kMinAUDSize;
 114     DCHECK_LE(next_access_unit_pos_, es_queue_->tail());
 115   }
 116   if (!FindAUD(&next_access_unit_pos_))
 117     return true;
 118
 119   // At this point, we know we have a full access unit.
 120   bool is_key_frame = false;
 121   int pps_id_for_access_unit = -1;
 122
 123   const uint8* es;
 124   int size;
 125   es_queue_->PeekAt(current_access_unit_pos_, &es, &size);
 126   int access_unit_size = base::checked_cast<int, int64>(
 127       next_access_unit_pos_ - current_access_unit_pos_);
 128   DCHECK_LE(access_unit_size, size);
 129   h264_parser_->SetStream(es, access_unit_size);
 130
 131   while (true) {
 132     bool is_eos = false;
 133     H264NALU nalu;
 134     switch (h264_parser_->AdvanceToNextNALU(&nalu)) {
 135       case H264Parser::kOk:
 136         break;
 137       case H264Parser::kInvalidStream:
 138       case H264Parser::kUnsupportedStream:
 139         return false;
 140       case H264Parser::kEOStream:
 141         is_eos = true;
 142         break;
 143     }
 144     if (is_eos)
 145       break;
 146
 147     switch (nalu.nal_unit_type) {
 148       case H264NALU::kAUD: {
 149         DVLOG(LOG_LEVEL_ES) << "NALU: AUD";
 150         break;
 151       }
 152       case H264NALU::kSPS: {
 153         DVLOG(LOG_LEVEL_ES) << "NALU: SPS";
 154         int sps_id;
 155         if (h264_parser_->ParseSPS(&sps_id) != H264Parser::kOk)
 156           return false;
 157         break;
 158       }
 159       case H264NALU::kPPS: {
 160         DVLOG(LOG_LEVEL_ES) << "NALU: PPS";
 161         int pps_id;
 162         if (h264_parser_->ParsePPS(&pps_id) != H264Parser::kOk)
 163           return false;
 164         break;
 165       }
 166       case H264NALU::kIDRSlice:
 167       case H264NALU::kNonIDRSlice: {
 168         is_key_frame = (nalu.nal_unit_type == H264NALU::kIDRSlice);
 169         DVLOG(LOG_LEVEL_ES) << "NALU: slice IDR=" << is_key_frame;
 170         H264SliceHeader shdr;
 171         if (h264_parser_->ParseSliceHeader(nalu, &shdr) != H264Parser::kOk) {
 172           // Only accept an invalid SPS/PPS at the beginning when the stream
 173           // does not necessarily start with an SPS/PPS/IDR.
 174           // TODO(damienv): Should be able to differentiate a missing SPS/PPS
 175           // from a slice header parsing error.
 176           if (last_video_decoder_config_.IsValidConfig())
 177             return false;
 178         } else {
 179           pps_id_for_access_unit = shdr.pic_parameter_set_id;
 180         }
 181         break;
 182       }
 183       default: {
 184         DVLOG(LOG_LEVEL_ES) << "NALU: " << nalu.nal_unit_type;
 185       }
 186     }
 187   }
 188
 189   // Emit a frame and move the stream to the next AUD position.
 190   RCHECK(EmitFrame(current_access_unit_pos_, access_unit_size,
 191                    is_key_frame, pps_id_for_access_unit));
 192   current_access_unit_pos_ = next_access_unit_pos_;
 193   es_queue_->Trim(current_access_unit_pos_);
 194
 195   return true;
 196 }
 197
 198 bool EsParserH264::EmitFrame(int64 access_unit_pos, int access_unit_size,
 199                              bool is_key_frame, int pps_id) {
 200   // Get the access unit timing info.
 201   // Note: |current_timing_desc.pts| might be |kNoTimestamp()| at this point
 202   // if:
 203   // - the stream is not fully MPEG-2 compliant.
 204   // - or if the stream relies on H264 VUI parameters to compute the timestamps.
 205   //   See H.222 spec: section 2.7.5 "Conditional coding of timestamps".
 206   //   This part is not yet implemented in EsParserH264.
 207   // |es_adapter_| will take care of the missing timestamps.
 208   TimingDesc current_timing_desc = GetTimingDescriptor(access_unit_pos);
 209   DVLOG_IF(1, current_timing_desc.pts == kNoTimestamp())
 210       << "Missing timestamp";
 211
 212   // If only the PTS is provided, copy the PTS into the DTS.
 213   if (current_timing_desc.dts == kNoDecodeTimestamp()) {
 214     current_timing_desc.dts =
 215         DecodeTimestamp::FromPresentationTime(current_timing_desc.pts);
 216   }
 217
 218   // Update the video decoder configuration if needed.
 219   const H264PPS* pps = h264_parser_->GetPPS(pps_id);
 220   if (!pps) {
 221     // Only accept an invalid PPS at the beginning when the stream
 222     // does not necessarily start with an SPS/PPS/IDR.
 223     // In this case, the initial frames are conveyed to the upper layer with
 224     // an invalid VideoDecoderConfig and it's up to the upper layer
 225     // to process this kind of frame accordingly.
 226     if (last_video_decoder_config_.IsValidConfig())
 227       return false;
 228   } else {
 229     const H264SPS* sps = h264_parser_->GetSPS(pps->seq_parameter_set_id);
 230     if (!sps)
 231       return false;
 232     RCHECK(UpdateVideoDecoderConfig(sps));
 233   }
 234
 235   // Emit a frame.
 236   DVLOG(LOG_LEVEL_ES) << "Emit frame: stream_pos=" << current_access_unit_pos_
 237                       << " size=" << access_unit_size;
 238   int es_size;
 239   const uint8* es;
 240   es_queue_->PeekAt(current_access_unit_pos_, &es, &es_size);
 241   CHECK_GE(es_size, access_unit_size);
 242
 243   // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
 244   // type and allow multiple video tracks. See https://crbug.com/341581.
 245   scoped_refptr<StreamParserBuffer> stream_parser_buffer =
 246       StreamParserBuffer::CopyFrom(
 247           es,
 248           access_unit_size,
 249           is_key_frame,
 250           DemuxerStream::VIDEO,
 251           0);
 252   stream_parser_buffer->SetDecodeTimestamp(current_timing_desc.dts);
 253   stream_parser_buffer->set_timestamp(current_timing_desc.pts);
 254   return es_adapter_.OnNewBuffer(stream_parser_buffer);
 255 }
 256
 257 bool EsParserH264::UpdateVideoDecoderConfig(const H264SPS* sps) {
 258   // Set the SAR to 1 when not specified in the H264 stream.
 259   int sar_width = (sps->sar_width == 0) ? 1 : sps->sar_width;
 260   int sar_height = (sps->sar_height == 0) ? 1 : sps->sar_height;
 261
 262   // TODO(damienv): a MAP unit can be either 16 or 32 pixels.
 263   // although it's 16 pixels for progressive non MBAFF frames.
 264   gfx::Size coded_size((sps->pic_width_in_mbs_minus1 + 1) * 16,
 265                        (sps->pic_height_in_map_units_minus1 + 1) * 16);
 266   gfx::Rect visible_rect(
 267       sps->frame_crop_left_offset,
 268       sps->frame_crop_top_offset,
 269       (coded_size.width() - sps->frame_crop_right_offset) -
 270       sps->frame_crop_left_offset,
 271       (coded_size.height() - sps->frame_crop_bottom_offset) -
 272       sps->frame_crop_top_offset);
 273   if (visible_rect.width() <= 0 || visible_rect.height() <= 0)
 274     return false;
 275   gfx::Size natural_size(
 276       (visible_rect.width() * sar_width) / sar_height,
 277       visible_rect.height());
 278   if (natural_size.width() == 0)
 279     return false;
 280
 281   VideoDecoderConfig video_decoder_config(
 282       kCodecH264, VIDEO_CODEC_PROFILE_UNKNOWN, PIXEL_FORMAT_YV12,
 283       COLOR_SPACE_HD_REC709, coded_size, visible_rect, natural_size, NULL, 0,
 284       false);
 285
 286   if (!video_decoder_config.Matches(last_video_decoder_config_)) {
 287     DVLOG(1) << "Profile IDC: " << sps->profile_idc;
 288     DVLOG(1) << "Level IDC: " << sps->level_idc;
 289     DVLOG(1) << "Pic width: " << coded_size.width();
 290     DVLOG(1) << "Pic height: " << coded_size.height();
 291     DVLOG(1) << "log2_max_frame_num_minus4: "
 292              << sps->log2_max_frame_num_minus4;
 293     DVLOG(1) << "SAR: width=" << sps->sar_width
 294              << " height=" << sps->sar_height;
 295     last_video_decoder_config_ = video_decoder_config;
 296     es_adapter_.OnConfigChanged(video_decoder_config);
 297   }
 298
 299   return true;
 300 }
 301
 302 }  // namespace mp2t
 303 }  // namespace media