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/basictypes.h"
   8 #include "base/logging.h"
   9 #include "base/numerics/safe_conversions.h"
  10 #include "media/base/buffers.h"
  11 #include "media/base/stream_parser_buffer.h"
  12 #include "media/base/video_frame.h"
  13 #include "media/filters/h264_parser.h"
  14 #include "media/formats/common/offset_byte_queue.h"
  15 #include "media/formats/mp2t/es_adapter_video.h"
  16 #include "media/formats/mp2t/mp2t_common.h"
  17 #include "ui/gfx/geometry/rect.h"
  18 #include "ui/gfx/geometry/size.h"
  19
  20 namespace media {
  21 namespace mp2t {
  22
  23 // An AUD NALU is at least 4 bytes:
  24 // 3 bytes for the start code + 1 byte for the NALU type.
  25 const int kMinAUDSize = 4;
  26
  27 EsParserH264::EsParserH264(
  28     const NewVideoConfigCB& new_video_config_cb,
  29     const EmitBufferCB& emit_buffer_cb)
  30     : es_adapter_(new_video_config_cb, emit_buffer_cb),
  31       h264_parser_(new H264Parser()),
  32       current_access_unit_pos_(0),
  33       next_access_unit_pos_(0) {
  34 }
  35
  36 EsParserH264::~EsParserH264() {
  37 }
  38
  39 void EsParserH264::Flush() {
  40   DVLOG(1) << __FUNCTION__;
  41   if (!FindAUD(&current_access_unit_pos_))
  42     return;
  43
  44   // Simulate an additional AUD to force emitting the last access unit
  45   // which is assumed to be complete at this point.
  46   uint8 aud[] = { 0x00, 0x00, 0x01, 0x09 };
  47   es_queue_->Push(aud, sizeof(aud));
  48   ParseFromEsQueue();
  49
  50   es_adapter_.Flush();
  51 }
  52
  53 void EsParserH264::ResetInternal() {
  54   DVLOG(1) << __FUNCTION__;
  55   h264_parser_.reset(new H264Parser());
  56   current_access_unit_pos_ = 0;
  57   next_access_unit_pos_ = 0;
  58   last_video_decoder_config_ = VideoDecoderConfig();
  59   es_adapter_.Reset();
  60 }
  61
  62 bool EsParserH264::FindAUD(int64* stream_pos) {
  63   while (true) {
  64     const uint8* es;
  65     int size;
  66     es_queue_->PeekAt(*stream_pos, &es, &size);
  67
  68     // Find a start code and move the stream to the start code parser position.
  69     off_t start_code_offset;
  70     off_t start_code_size;
  71     bool start_code_found = H264Parser::FindStartCode(
  72         es, size, &start_code_offset, &start_code_size);
  73     *stream_pos += start_code_offset;
  74
  75     // No H264 start code found or NALU type not available yet.
  76     if (!start_code_found || start_code_offset + start_code_size >= size)
  77       return false;
  78
  79     // Exit the parser loop when an AUD is found.
  80     // Note: NALU header for an AUD:
  81     // - nal_ref_idc must be 0
  82     // - nal_unit_type must be H264NALU::kAUD
  83     if (es[start_code_offset + start_code_size] == H264NALU::kAUD)
  84       break;
  85
  86     // The current NALU is not an AUD, skip the start code
  87     // and continue parsing the stream.
  88     *stream_pos += start_code_size;
  89   }
  90
  91   return true;
  92 }
  93
  94 bool EsParserH264::ParseFromEsQueue() {
  95   DCHECK_LE(es_queue_->head(), current_access_unit_pos_);
  96   DCHECK_LE(current_access_unit_pos_, next_access_unit_pos_);
  97   DCHECK_LE(next_access_unit_pos_, es_queue_->tail());
  98
  99   // Find the next AUD located at or after |current_access_unit_pos_|. This is
 100   // needed since initially |current_access_unit_pos_| might not point to
 101   // an AUD.
 102   // Discard all the data before the updated |current_access_unit_pos_|
 103   // since it won't be used again.
 104   bool aud_found = FindAUD(&current_access_unit_pos_);
 105   es_queue_->Trim(current_access_unit_pos_);
 106   if (next_access_unit_pos_ < current_access_unit_pos_)
 107     next_access_unit_pos_ = current_access_unit_pos_;
 108
 109   // Resume parsing later if no AUD was found.
 110   if (!aud_found)
 111     return true;
 112
 113   // Find the next AUD to make sure we have a complete access unit.
 114   if (next_access_unit_pos_ < current_access_unit_pos_ + kMinAUDSize) {
 115     next_access_unit_pos_ = current_access_unit_pos_ + kMinAUDSize;
 116     DCHECK_LE(next_access_unit_pos_, es_queue_->tail());
 117   }
 118   if (!FindAUD(&next_access_unit_pos_))
 119     return true;
 120
 121   // At this point, we know we have a full access unit.
 122   bool is_key_frame = false;
 123   int pps_id_for_access_unit = -1;
 124
 125   const uint8* es;
 126   int size;
 127   es_queue_->PeekAt(current_access_unit_pos_, &es, &size);
 128   int access_unit_size = base::checked_cast<int, int64>(
 129       next_access_unit_pos_ - current_access_unit_pos_);
 130   DCHECK_LE(access_unit_size, size);
 131   h264_parser_->SetStream(es, access_unit_size);
 132
 133   while (true) {
 134     bool is_eos = false;
 135     H264NALU nalu;
 136     switch (h264_parser_->AdvanceToNextNALU(&nalu)) {
 137       case H264Parser::kOk:
 138         break;
 139       case H264Parser::kInvalidStream:
 140       case H264Parser::kUnsupportedStream:
 141         return false;
 142       case H264Parser::kEOStream:
 143         is_eos = true;
 144         break;
 145     }
 146     if (is_eos)
 147       break;
 148
 149     switch (nalu.nal_unit_type) {
 150       case H264NALU::kAUD: {
 151         DVLOG(LOG_LEVEL_ES) << "NALU: AUD";
 152         break;
 153       }
 154       case H264NALU::kSPS: {
 155         DVLOG(LOG_LEVEL_ES) << "NALU: SPS";
 156         int sps_id;
 157         if (h264_parser_->ParseSPS(&sps_id) != H264Parser::kOk)
 158           return false;
 159         break;
 160       }
 161       case H264NALU::kPPS: {
 162         DVLOG(LOG_LEVEL_ES) << "NALU: PPS";
 163         int pps_id;
 164         if (h264_parser_->ParsePPS(&pps_id) != H264Parser::kOk)
 165           return false;
 166         break;
 167       }
 168       case H264NALU::kIDRSlice:
 169       case H264NALU::kNonIDRSlice: {
 170         is_key_frame = (nalu.nal_unit_type == H264NALU::kIDRSlice);
 171         DVLOG(LOG_LEVEL_ES) << "NALU: slice IDR=" << is_key_frame;
 172         H264SliceHeader shdr;
 173         if (h264_parser_->ParseSliceHeader(nalu, &shdr) != H264Parser::kOk) {
 174           // Only accept an invalid SPS/PPS at the beginning when the stream
 175           // does not necessarily start with an SPS/PPS/IDR.
 176           // TODO(damienv): Should be able to differentiate a missing SPS/PPS
 177           // from a slice header parsing error.
 178           if (last_video_decoder_config_.IsValidConfig())
 179             return false;
 180         } else {
 181           pps_id_for_access_unit = shdr.pic_parameter_set_id;
 182         }
 183         break;
 184       }
 185       default: {
 186         DVLOG(LOG_LEVEL_ES) << "NALU: " << nalu.nal_unit_type;
 187       }
 188     }
 189   }
 190
 191   // Emit a frame and move the stream to the next AUD position.
 192   RCHECK(EmitFrame(current_access_unit_pos_, access_unit_size,
 193                    is_key_frame, pps_id_for_access_unit));
 194   current_access_unit_pos_ = next_access_unit_pos_;
 195   es_queue_->Trim(current_access_unit_pos_);
 196
 197   return true;
 198 }
 199
 200 bool EsParserH264::EmitFrame(int64 access_unit_pos, int access_unit_size,
 201                              bool is_key_frame, int pps_id) {
 202   // Get the access unit timing info.
 203   // Note: |current_timing_desc.pts| might be |kNoTimestamp()| at this point
 204   // if:
 205   // - the stream is not fully MPEG-2 compliant.
 206   // - or if the stream relies on H264 VUI parameters to compute the timestamps.
 207   //   See H.222 spec: section 2.7.5 "Conditional coding of timestamps".
 208   //   This part is not yet implemented in EsParserH264.
 209   // |es_adapter_| will take care of the missing timestamps.
 210   TimingDesc current_timing_desc = GetTimingDescriptor(access_unit_pos);
 211   DVLOG_IF(1, current_timing_desc.pts == kNoTimestamp())
 212       << "Missing timestamp";
 213
 214   // If only the PTS is provided, copy the PTS into the DTS.
 215   if (current_timing_desc.dts == kNoDecodeTimestamp()) {
 216     current_timing_desc.dts =
 217         DecodeTimestamp::FromPresentationTime(current_timing_desc.pts);
 218   }
 219
 220   // Update the video decoder configuration if needed.
 221   const H264PPS* pps = h264_parser_->GetPPS(pps_id);
 222   if (!pps) {
 223     // Only accept an invalid PPS at the beginning when the stream
 224     // does not necessarily start with an SPS/PPS/IDR.
 225     // In this case, the initial frames are conveyed to the upper layer with
 226     // an invalid VideoDecoderConfig and it's up to the upper layer
 227     // to process this kind of frame accordingly.
 228     if (last_video_decoder_config_.IsValidConfig())
 229       return false;
 230   } else {
 231     const H264SPS* sps = h264_parser_->GetSPS(pps->seq_parameter_set_id);
 232     if (!sps)
 233       return false;
 234     RCHECK(UpdateVideoDecoderConfig(sps));
 235   }
 236
 237   // Emit a frame.
 238   DVLOG(LOG_LEVEL_ES) << "Emit frame: stream_pos=" << current_access_unit_pos_
 239                       << " size=" << access_unit_size;
 240   int es_size;
 241   const uint8* es;
 242   es_queue_->PeekAt(current_access_unit_pos_, &es, &es_size);
 243   CHECK_GE(es_size, access_unit_size);
 244
 245   // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
 246   // type and allow multiple video tracks. See https://crbug.com/341581.
 247   scoped_refptr<StreamParserBuffer> stream_parser_buffer =
 248       StreamParserBuffer::CopyFrom(
 249           es,
 250           access_unit_size,
 251           is_key_frame,
 252           DemuxerStream::VIDEO,
 253           0);
 254   stream_parser_buffer->SetDecodeTimestamp(current_timing_desc.dts);
 255   stream_parser_buffer->set_timestamp(current_timing_desc.pts);
 256   return es_adapter_.OnNewBuffer(stream_parser_buffer);
 257 }
 258
 259 bool EsParserH264::UpdateVideoDecoderConfig(const H264SPS* sps) {
 260   // Set the SAR to 1 when not specified in the H264 stream.
 261   int sar_width = (sps->sar_width == 0) ? 1 : sps->sar_width;
 262   int sar_height = (sps->sar_height == 0) ? 1 : sps->sar_height;
 263
 264   // TODO(damienv): a MAP unit can be either 16 or 32 pixels.
 265   // although it's 16 pixels for progressive non MBAFF frames.
 266   gfx::Size coded_size((sps->pic_width_in_mbs_minus1 + 1) * 16,
 267                        (sps->pic_height_in_map_units_minus1 + 1) * 16);
 268   gfx::Rect visible_rect(
 269       sps->frame_crop_left_offset,
 270       sps->frame_crop_top_offset,
 271       (coded_size.width() - sps->frame_crop_right_offset) -
 272       sps->frame_crop_left_offset,
 273       (coded_size.height() - sps->frame_crop_bottom_offset) -
 274       sps->frame_crop_top_offset);
 275   if (visible_rect.width() <= 0 || visible_rect.height() <= 0)
 276     return false;
 277   gfx::Size natural_size(
 278       (visible_rect.width() * sar_width) / sar_height,
 279       visible_rect.height());
 280   if (natural_size.width() == 0)
 281     return false;
 282
 283   VideoDecoderConfig video_decoder_config(
 284       kCodecH264,
 285       VIDEO_CODEC_PROFILE_UNKNOWN,
 286       VideoFrame::YV12,
 287       coded_size,
 288       visible_rect,
 289       natural_size,
 290       NULL, 0,
 291       false);
 292
 293   if (!video_decoder_config.Matches(last_video_decoder_config_)) {
 294     DVLOG(1) << "Profile IDC: " << sps->profile_idc;
 295     DVLOG(1) << "Level IDC: " << sps->level_idc;
 296     DVLOG(1) << "Pic width: " << coded_size.width();
 297     DVLOG(1) << "Pic height: " << coded_size.height();
 298     DVLOG(1) << "log2_max_frame_num_minus4: "
 299              << sps->log2_max_frame_num_minus4;
 300     DVLOG(1) << "SAR: width=" << sps->sar_width
 301              << " height=" << sps->sar_height;
 302     last_video_decoder_config_ = video_decoder_config;
 303     es_adapter_.OnConfigChanged(video_decoder_config);
 304   }
 305
 306   return true;
 307 }
 308
 309 }  // namespace mp2t
 310 }  // namespace media