Properly handle AVC profiles with SPS extension
Part of #755 to improve DASH spec compliance.
This commit is contained in:
koln67
GitHub
parent
cbc251c621
commit
caa47e374d
|
|
@ -15,18 +15,18 @@
|
|||
<AdaptationSet id="1" contentType="video" maxWidth="1280" maxHeight="720" frameRate="30000/1001" subsegmentAlignment="true" par="16:9">
|
||||
<SupplementalProperty schemeIdUri="urn:mpeg:dash:adaptation-set-switching:2016" value="0"/>
|
||||
<Role schemeIdUri="urn:mpeg:dash:role:2011" value="main"/>
|
||||
<Representation id="1" bandwidth="973483" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360">
|
||||
<BaseURL>bear-640x360-video.mp4</BaseURL>
|
||||
<SegmentBase indexRange="859-926" timescale="30000">
|
||||
<Initialization range="0-858"/>
|
||||
</SegmentBase>
|
||||
</Representation>
|
||||
<Representation id="2" bandwidth="2627285" codecs="avc1.64001f" mimeType="video/mp4" sar="1:1" width="1280" height="720">
|
||||
<Representation id="1" bandwidth="2627285" codecs="avc1.64001f" mimeType="video/mp4" sar="1:1" width="1280" height="720">
|
||||
<BaseURL>bear-1280x720-video.mp4</BaseURL>
|
||||
<SegmentBase indexRange="858-925" timescale="30000">
|
||||
<Initialization range="0-857"/>
|
||||
</SegmentBase>
|
||||
</Representation>
|
||||
<Representation id="2" bandwidth="973483" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360">
|
||||
<BaseURL>bear-640x360-video.mp4</BaseURL>
|
||||
<SegmentBase indexRange="859-926" timescale="30000">
|
||||
<Initialization range="0-858"/>
|
||||
</SegmentBase>
|
||||
</Representation>
|
||||
</AdaptationSet>
|
||||
<AdaptationSet id="2" contentType="audio" subsegmentAlignment="true">
|
||||
<Representation id="3" bandwidth="133334" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
|
||||
|
|
|
|||
|
|
@ -4,14 +4,14 @@
|
|||
#EXT-X-TARGETDURATION:2
|
||||
#EXT-X-PLAYLIST-TYPE:VOD
|
||||
#EXT-X-I-FRAMES-ONLY
|
||||
#EXT-X-MAP:URI="bear-640x360-ac3-video.mp4",BYTERANGE="859@0"
|
||||
#EXT-X-MAP:URI="bear-640x360-ac3-video.mp4",BYTERANGE="863@0"
|
||||
#EXTINF:1.001,
|
||||
#EXT-X-BYTERANGE:15581@927
|
||||
#EXT-X-BYTERANGE:15581@931
|
||||
bear-640x360-ac3-video.mp4
|
||||
#EXTINF:1.001,
|
||||
#EXT-X-BYTERANGE:18221@100240
|
||||
#EXT-X-BYTERANGE:18221@100244
|
||||
bear-640x360-ac3-video.mp4
|
||||
#EXTINF:0.734,
|
||||
#EXT-X-BYTERANGE:19663@222047
|
||||
#EXT-X-BYTERANGE:19663@222051
|
||||
bear-640x360-ac3-video.mp4
|
||||
#EXT-X-ENDLIST
|
||||
|
|
|
|||
|
|
@ -3,9 +3,9 @@
|
|||
## Generated with https://github.com/google/shaka-packager version <tag>-<hash>-<test>
|
||||
#EXT-X-TARGETDURATION:2
|
||||
#EXT-X-PLAYLIST-TYPE:VOD
|
||||
#EXT-X-MAP:URI="bear-640x360-ac3-video.mp4",BYTERANGE="859@0"
|
||||
#EXT-X-MAP:URI="bear-640x360-ac3-video.mp4",BYTERANGE="863@0"
|
||||
#EXTINF:1.001,
|
||||
#EXT-X-BYTERANGE:99313@927
|
||||
#EXT-X-BYTERANGE:99313@931
|
||||
bear-640x360-ac3-video.mp4
|
||||
#EXTINF:1.001,
|
||||
#EXT-X-BYTERANGE:121807
|
||||
|
|
|
|||
Binary file not shown.
|
|
@ -2,18 +2,8 @@
|
|||
<!--Generated with https://github.com/google/shaka-packager version <tag>-<hash>-<test>-->
|
||||
<MPD xmlns="urn:mpeg:dash:schema:mpd:2011" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="urn:mpeg:dash:schema:mpd:2011 DASH-MPD.xsd" profiles="urn:mpeg:dash:profile:isoff-live:2011" minBufferTime="PT2S" type="dynamic" publishTime="some_time" availabilityStartTime="some_time" minimumUpdatePeriod="PT5S" timeShiftBufferDepth="PT1800S">
|
||||
<Period id="0" start="PT0S">
|
||||
<AdaptationSet id="0" contentType="text" segmentAlignment="true">
|
||||
<Role schemeIdUri="urn:mpeg:dash:role:2011" value="subtitle"/>
|
||||
<Representation id="0" bandwidth="2024" codecs="wvtt" mimeType="application/mp4">
|
||||
<SegmentTemplate timescale="1000" initialization="bear-english-text-init.mp4" media="bear-english-text-$Number$.m4s" startNumber="1">
|
||||
<SegmentTimeline>
|
||||
<S t="0" d="1000" r="4"/>
|
||||
</SegmentTimeline>
|
||||
</SegmentTemplate>
|
||||
</Representation>
|
||||
</AdaptationSet>
|
||||
<AdaptationSet id="1" contentType="video" width="640" height="360" frameRate="30000/1001" segmentAlignment="true" par="16:9">
|
||||
<Representation id="1" bandwidth="974122" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1">
|
||||
<AdaptationSet id="0" contentType="video" width="640" height="360" frameRate="30000/1001" segmentAlignment="true" par="16:9">
|
||||
<Representation id="0" bandwidth="974122" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1">
|
||||
<SegmentTemplate timescale="30000" initialization="bear-640x360-video-init.mp4" media="bear-640x360-video-$Number$.m4s" startNumber="1">
|
||||
<SegmentTimeline>
|
||||
<S t="0" d="30030" r="1"/>
|
||||
|
|
@ -22,6 +12,16 @@
|
|||
</SegmentTemplate>
|
||||
</Representation>
|
||||
</AdaptationSet>
|
||||
<AdaptationSet id="1" contentType="text" segmentAlignment="true">
|
||||
<Role schemeIdUri="urn:mpeg:dash:role:2011" value="subtitle"/>
|
||||
<Representation id="1" bandwidth="2024" codecs="wvtt" mimeType="application/mp4">
|
||||
<SegmentTemplate timescale="1000" initialization="bear-english-text-init.mp4" media="bear-english-text-$Number$.m4s" startNumber="1">
|
||||
<SegmentTimeline>
|
||||
<S t="0" d="1000" r="4"/>
|
||||
</SegmentTimeline>
|
||||
</SegmentTemplate>
|
||||
</Representation>
|
||||
</AdaptationSet>
|
||||
<AdaptationSet id="2" contentType="audio" segmentAlignment="true">
|
||||
<Representation id="2" bandwidth="133929" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
|
||||
<AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
|
||||
|
|
|
|||
|
|
@ -7,8 +7,8 @@
|
|||
<ContentProtection schemeIdUri="urn:uuid:1077efec-c0b2-4d02-ace3-3c1e52e2fb4b">
|
||||
<cenc:pssh>AAAANHBzc2gBAAAAEHfv7MCyTQKs4zweUuL7SwAAAAExMjM0NTY3ODkwMTIzNDU2AAAAAA==</cenc:pssh>
|
||||
</ContentProtection>
|
||||
<Representation id="0" bandwidth="978382" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360">
|
||||
<SegmentTemplate timescale="30000" initialization="bear-640x360-video-init.mp4" media="bear-640x360-video-$Number$.m4s" startNumber="1">
|
||||
<Representation id="0" bandwidth="2632184" codecs="avc1.64001f" mimeType="video/mp4" sar="1:1" width="1280" height="720">
|
||||
<SegmentTemplate timescale="30000" initialization="bear-1280x720-video-init.mp4" media="bear-1280x720-video-$Number$.m4s" startNumber="1">
|
||||
<SegmentTimeline>
|
||||
<S t="0" d="30030" r="1"/>
|
||||
<S t="60060" d="22022"/>
|
||||
|
|
@ -23,8 +23,8 @@
|
|||
</SegmentTimeline>
|
||||
</SegmentTemplate>
|
||||
</Representation>
|
||||
<Representation id="2" bandwidth="2632184" codecs="avc1.64001f" mimeType="video/mp4" sar="1:1" width="1280" height="720">
|
||||
<SegmentTemplate timescale="30000" initialization="bear-1280x720-video-init.mp4" media="bear-1280x720-video-$Number$.m4s" startNumber="1">
|
||||
<Representation id="2" bandwidth="978382" codecs="avc1.64001e" mimeType="video/mp4" sar="1:1" width="640" height="360">
|
||||
<SegmentTemplate timescale="30000" initialization="bear-640x360-video-init.mp4" media="bear-640x360-video-$Number$.m4s" startNumber="1">
|
||||
<SegmentTimeline>
|
||||
<S t="0" d="30030" r="1"/>
|
||||
<S t="60060" d="22022"/>
|
||||
|
|
@ -37,13 +37,13 @@
|
|||
<ContentProtection schemeIdUri="urn:uuid:1077efec-c0b2-4d02-ace3-3c1e52e2fb4b">
|
||||
<cenc:pssh>AAAANHBzc2gBAAAAEHfv7MCyTQKs4zweUuL7SwAAAAExMjM0NTY3ODkwMTIzNDU2AAAAAA==</cenc:pssh>
|
||||
</ContentProtection>
|
||||
<Representation id="3" bandwidth="134272" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
|
||||
<Representation id="3" bandwidth="134015" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
|
||||
<AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
|
||||
<SegmentTemplate timescale="44100" initialization="bear-640x360-audio-init.mp4" media="bear-640x360-audio-$Number$.m4s" startNumber="1">
|
||||
<SegmentTemplate timescale="44100" initialization="bear-320x180-audio-init.mp4" media="bear-320x180-audio-$Number$.m4s" startNumber="1">
|
||||
<SegmentTimeline>
|
||||
<S t="0" d="45056"/>
|
||||
<S t="45056" d="44032"/>
|
||||
<S t="89088" d="31744"/>
|
||||
<S t="89088" d="33792"/>
|
||||
</SegmentTimeline>
|
||||
</SegmentTemplate>
|
||||
</Representation>
|
||||
|
|
@ -57,13 +57,13 @@
|
|||
</SegmentTimeline>
|
||||
</SegmentTemplate>
|
||||
</Representation>
|
||||
<Representation id="5" bandwidth="134015" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
|
||||
<Representation id="5" bandwidth="134272" codecs="mp4a.40.2" mimeType="audio/mp4" audioSamplingRate="44100">
|
||||
<AudioChannelConfiguration schemeIdUri="urn:mpeg:dash:23003:3:audio_channel_configuration:2011" value="2"/>
|
||||
<SegmentTemplate timescale="44100" initialization="bear-320x180-audio-init.mp4" media="bear-320x180-audio-$Number$.m4s" startNumber="1">
|
||||
<SegmentTemplate timescale="44100" initialization="bear-640x360-audio-init.mp4" media="bear-640x360-audio-$Number$.m4s" startNumber="1">
|
||||
<SegmentTimeline>
|
||||
<S t="0" d="45056"/>
|
||||
<S t="45056" d="44032"/>
|
||||
<S t="89088" d="33792"/>
|
||||
<S t="89088" d="31744"/>
|
||||
</SegmentTimeline>
|
||||
</SegmentTemplate>
|
||||
</Representation>
|
||||
|
|
|
|||
|
|
@ -82,6 +82,30 @@ bool AVCDecoderConfigurationRecord::ParseInternal() {
|
|||
AddNalu(nalu);
|
||||
}
|
||||
|
||||
if (profile_indication_ == 100 || profile_indication_ == 110 ||
|
||||
profile_indication_ == 122 || profile_indication_ == 144) {
|
||||
|
||||
uint8_t sps_ext_count;
|
||||
if (!reader.Read1(&chroma_format_) || !reader.Read1(&bit_depth_luma_minus8_) ||
|
||||
!reader.Read1(&bit_depth_chroma_minus8_) || !reader.Read1(&sps_ext_count)) {
|
||||
LOG(WARNING) << "Insufficient bits in bitstream for given AVC profile";
|
||||
return true;
|
||||
}
|
||||
chroma_format_ &= 0x3;
|
||||
bit_depth_luma_minus8_ &= 0x7;
|
||||
bit_depth_chroma_minus8_ &= 0x7;
|
||||
for (uint8_t i = 0; i < sps_ext_count; i++) {
|
||||
uint16_t size = 0;
|
||||
RCHECK(reader.Read2(&size));
|
||||
const uint8_t* nalu_data = reader.data() + reader.pos();
|
||||
RCHECK(reader.SkipBytes(size));
|
||||
|
||||
Nalu nalu;
|
||||
RCHECK(nalu.Initialize(Nalu::kH264, nalu_data, size));
|
||||
RCHECK(nalu.type() == Nalu::H264_SPSExtension);
|
||||
AddNalu(nalu);
|
||||
}
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -35,6 +35,9 @@ class AVCDecoderConfigurationRecord : public DecoderConfigurationRecord {
|
|||
uint32_t coded_height() const { return coded_height_; }
|
||||
uint32_t pixel_width() const { return pixel_width_; }
|
||||
uint32_t pixel_height() const { return pixel_height_; }
|
||||
uint8_t chroma_format() const { return chroma_format_; }
|
||||
uint8_t bit_depth_luma_minus8() const { return bit_depth_luma_minus8_; }
|
||||
uint8_t bit_depth_chroma_minus8() const { return bit_depth_chroma_minus8_; }
|
||||
|
||||
/// Static version of GetCodecString.
|
||||
/// @return The codec string.
|
||||
|
|
@ -56,6 +59,12 @@ class AVCDecoderConfigurationRecord : public DecoderConfigurationRecord {
|
|||
uint32_t coded_height_ = 0;
|
||||
uint32_t pixel_width_ = 0;
|
||||
uint32_t pixel_height_ = 0;
|
||||
|
||||
// Only should be present for special case profile values.
|
||||
// Refer to ISO/IEC 14496-15 Section 5.3.3.1.1.
|
||||
uint8_t chroma_format_ = 0;
|
||||
uint8_t bit_depth_luma_minus8_ = 0;
|
||||
uint8_t bit_depth_chroma_minus8_ = 0;
|
||||
|
||||
DISALLOW_COPY_AND_ASSIGN(AVCDecoderConfigurationRecord);
|
||||
};
|
||||
|
|
|
|||
|
|
@ -45,11 +45,59 @@ TEST(AVCDecoderConfigurationRecordTest, Success) {
|
|||
EXPECT_EQ(8u, avc_config.pixel_width());
|
||||
EXPECT_EQ(9u, avc_config.pixel_height());
|
||||
EXPECT_EQ(0u, avc_config.transfer_characteristics());
|
||||
EXPECT_EQ(0u, avc_config.chroma_format());
|
||||
EXPECT_EQ(0u, avc_config.bit_depth_luma_minus8());
|
||||
EXPECT_EQ(0u, avc_config.bit_depth_chroma_minus8());
|
||||
|
||||
EXPECT_EQ("avc1.64001e", avc_config.GetCodecString(FOURCC_avc1));
|
||||
EXPECT_EQ("avc3.64001e", avc_config.GetCodecString(FOURCC_avc3));
|
||||
}
|
||||
|
||||
TEST(AVCDecoderConfigurationRecordTest, SuccessWithSPSExtension) {
|
||||
// clang-format off
|
||||
const uint8_t kAvcDecoderConfigurationData[] = {
|
||||
0x01, // version
|
||||
0x64, // profile_indication
|
||||
0x00, // profile_compatibility
|
||||
0x1E, // avc_level
|
||||
0xFF, // Least significant 3 bits is length_size_minus_one
|
||||
0xE1, // Least significant 5 bits is num_sps
|
||||
// sps 1
|
||||
0x00, 0x1D, // size
|
||||
0x67, 0x64, 0x00, 0x1E, 0xAC, 0xD9, 0x40, 0xB4, 0x2F, 0xF9, 0x7F, 0xF0,
|
||||
0x00, 0x80, 0x00, 0x91, 0x00, 0x00, 0x03, 0x03, 0xE9, 0x00, 0x00, 0xEA,
|
||||
0x60, 0x0F, 0x16, 0x2D, 0x96,
|
||||
0x01, // num_pps
|
||||
0x00, 0x06, // size
|
||||
0x68, 0xEB, 0xE3, 0xCB, 0x22, 0xC0,
|
||||
0xFC, // Least significant 2 bits is chroma_format
|
||||
0xF9, // Least significant 3 bits is bit_depth_luma_minus8
|
||||
0xFF, // Least significant 3 bits is bit_depth_chroma_minus8
|
||||
0x01, // num_sps_ext
|
||||
0x00, 0x05, // size
|
||||
0x6D, 0x33, 0x01, 0x57, 0x78
|
||||
};
|
||||
// clang-format on
|
||||
|
||||
AVCDecoderConfigurationRecord avc_config;
|
||||
ASSERT_TRUE(avc_config.Parse(kAvcDecoderConfigurationData,
|
||||
arraysize(kAvcDecoderConfigurationData)));
|
||||
|
||||
EXPECT_EQ(1u, avc_config.version());
|
||||
EXPECT_EQ(0x64, avc_config.profile_indication());
|
||||
EXPECT_EQ(0u, avc_config.profile_compatibility());
|
||||
EXPECT_EQ(0x1E, avc_config.avc_level());
|
||||
EXPECT_EQ(4u, avc_config.nalu_length_size());
|
||||
EXPECT_EQ(720u, avc_config.coded_width());
|
||||
EXPECT_EQ(360u, avc_config.coded_height());
|
||||
EXPECT_EQ(8u, avc_config.pixel_width());
|
||||
EXPECT_EQ(9u, avc_config.pixel_height());
|
||||
EXPECT_EQ(0u, avc_config.transfer_characteristics());
|
||||
EXPECT_EQ(0u, avc_config.chroma_format());
|
||||
EXPECT_EQ(1u, avc_config.bit_depth_luma_minus8());
|
||||
EXPECT_EQ(7u, avc_config.bit_depth_chroma_minus8());
|
||||
}
|
||||
|
||||
TEST(AVCDecoderConfigurationRecordTest, SuccessWithTransferCharacteristics) {
|
||||
// clang-format off
|
||||
const uint8_t kAvcDecoderConfigurationData[] = {
|
||||
|
|
|
|||
|
|
@ -15,6 +15,21 @@
|
|||
namespace shaka {
|
||||
namespace media {
|
||||
|
||||
namespace {
|
||||
// utility helper function to get an sps
|
||||
const H264Sps* ParseSpsFromBytes(const std::vector<uint8_t> sps,
|
||||
H264Parser* parser) {
|
||||
Nalu nalu;
|
||||
if (!nalu.Initialize(Nalu::kH264, sps.data(), sps.size()))
|
||||
return nullptr;
|
||||
|
||||
int sps_id = 0;
|
||||
if (parser->ParseSps(nalu, &sps_id) != H264Parser::kOk)
|
||||
return nullptr;
|
||||
return parser->GetSps(sps_id);
|
||||
}
|
||||
} // namespace
|
||||
|
||||
H264ByteToUnitStreamConverter::H264ByteToUnitStreamConverter()
|
||||
: H26xByteToUnitStreamConverter(Nalu::kH264) {}
|
||||
|
||||
|
|
@ -27,15 +42,14 @@ H264ByteToUnitStreamConverter::~H264ByteToUnitStreamConverter() {}
|
|||
bool H264ByteToUnitStreamConverter::GetDecoderConfigurationRecord(
|
||||
std::vector<uint8_t>* decoder_config) const {
|
||||
DCHECK(decoder_config);
|
||||
|
||||
if ((last_sps_.size() < 4) || last_pps_.empty()) {
|
||||
// No data available to construct AVCDecoderConfigurationRecord.
|
||||
return false;
|
||||
}
|
||||
|
||||
// Construct an AVCDecoderConfigurationRecord containing a single SPS and a
|
||||
// single PPS NALU. Please refer to ISO/IEC 14496-15 for format specifics.
|
||||
BufferWriter buffer(last_sps_.size() + last_pps_.size() + 11);
|
||||
// Construct an AVCDecoderConfigurationRecord containing a single SPS, a
|
||||
// single PPS, and if available, a single SPS Extension NALU.
|
||||
// Please refer to ISO/IEC 14496-15 for format specifics.
|
||||
BufferWriter buffer;
|
||||
uint8_t version(1);
|
||||
buffer.AppendInt(version);
|
||||
buffer.AppendInt(last_sps_[1]);
|
||||
|
|
@ -51,6 +65,32 @@ bool H264ByteToUnitStreamConverter::GetDecoderConfigurationRecord(
|
|||
buffer.AppendInt(num_pps);
|
||||
buffer.AppendInt(static_cast<uint16_t>(last_pps_.size()));
|
||||
buffer.AppendVector(last_pps_);
|
||||
// handle profile special cases, refer to ISO/IEC 14496-15 Section 5.3.3.1.2
|
||||
uint8_t profile_indication = last_sps_[1];
|
||||
if (profile_indication == 100 || profile_indication == 110 ||
|
||||
profile_indication == 122 || profile_indication == 144) {
|
||||
|
||||
H264Parser parser;
|
||||
const H264Sps* sps = ParseSpsFromBytes(last_sps_, &parser);
|
||||
if (sps == nullptr)
|
||||
return false;
|
||||
|
||||
uint8_t reserved_chroma_format = 0xfc | (sps->chroma_format_idc);
|
||||
buffer.AppendInt(reserved_chroma_format);
|
||||
uint8_t reserved_bit_depth_luma_minus8 = 0xf8 | sps->bit_depth_luma_minus8;
|
||||
buffer.AppendInt(reserved_bit_depth_luma_minus8);
|
||||
uint8_t reserved_bit_depth_chroma_minus8 = 0xf8 | sps->bit_depth_chroma_minus8;
|
||||
buffer.AppendInt(reserved_bit_depth_chroma_minus8);
|
||||
|
||||
if (last_sps_ext_.empty()) {
|
||||
uint8_t num_sps_ext(0);
|
||||
buffer.AppendInt(num_sps_ext);
|
||||
} else {
|
||||
uint8_t num_sps_ext(1);
|
||||
buffer.AppendInt(num_sps_ext);
|
||||
buffer.AppendVector(last_sps_ext_);
|
||||
}
|
||||
}
|
||||
|
||||
buffer.SwapBuffer(decoder_config);
|
||||
return true;
|
||||
|
|
@ -70,6 +110,12 @@ bool H264ByteToUnitStreamConverter::ProcessNalu(const Nalu& nalu) {
|
|||
// Grab SPS NALU.
|
||||
last_sps_.assign(nalu_ptr, nalu_ptr + nalu_size);
|
||||
return strip_parameter_set_nalus();
|
||||
case Nalu::H264_SPSExtension:
|
||||
if (strip_parameter_set_nalus())
|
||||
WarnIfNotMatch(nalu.type(), nalu_ptr, nalu_size, last_sps_ext_);
|
||||
// Grab SPSExtension NALU.
|
||||
last_sps_ext_.assign(nalu_ptr, nalu_ptr + nalu_size);
|
||||
return strip_parameter_set_nalus();
|
||||
case Nalu::H264_PPS:
|
||||
if (strip_parameter_set_nalus())
|
||||
WarnIfNotMatch(nalu.type(), nalu_ptr, nalu_size, last_pps_);
|
||||
|
|
|
|||
|
|
@ -42,6 +42,7 @@ class H264ByteToUnitStreamConverter : public H26xByteToUnitStreamConverter {
|
|||
|
||||
std::vector<uint8_t> last_sps_;
|
||||
std::vector<uint8_t> last_pps_;
|
||||
std::vector<uint8_t> last_sps_ext_;
|
||||
|
||||
DISALLOW_COPY_AND_ASSIGN(H264ByteToUnitStreamConverter);
|
||||
};
|
||||
|
|
|
|||
|
|
@ -79,5 +79,64 @@ TEST(H264ByteToUnitStreamConverter, ConversionFailure) {
|
|||
EXPECT_FALSE(converter.GetDecoderConfigurationRecord(&decoder_config));
|
||||
}
|
||||
|
||||
TEST(H264ByteToUnitStreamConverter, NaluConversionWithSpsExtension) {
|
||||
const uint8_t kByteStreamWithSpsExtension[] = {
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
0x09, // AUD Type
|
||||
0xF0, // Primary pic type
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// Some SPS Data
|
||||
0x67, 0x64, 0x00, 0x1E, 0xAC, 0xD9, 0x40, 0xB4, 0x2F,
|
||||
0xF9, 0x7F, 0xF0, 0x00, 0x80, 0x00, 0x91, 0x00, 0x00,
|
||||
0x03, 0x03, 0xE9, 0x00, 0x00, 0xEA, 0x60, 0x0F, 0x16,
|
||||
0x2D, 0x96,
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// Some PPS Data
|
||||
0x68, 0xFE, 0xFD, 0xFC, 0xFB, 0x11, 0x12, 0x13, 0x14, 0x15,
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// Some SPS Extension data
|
||||
0x6D, 0x33, 0x01, 0x57, 0x78,
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// The input NALU
|
||||
0x06, // NALU type
|
||||
0xFD, 0x78, 0xA4, 0xC3, 0x82, 0x62, 0x11, 0x29, 0x77,
|
||||
};
|
||||
std::vector<uint8_t> byte_stream_with_sps_extension(
|
||||
std::begin(kByteStreamWithSpsExtension),
|
||||
std::end(kByteStreamWithSpsExtension));
|
||||
|
||||
H264ByteToUnitStreamConverter converter(
|
||||
H26xStreamFormat::kNalUnitStreamWithoutParameterSetNalus);
|
||||
|
||||
std::vector<uint8_t> unit_stream;
|
||||
ASSERT_TRUE(converter.ConvertByteStreamToNalUnitStream(
|
||||
byte_stream_with_sps_extension.data(),
|
||||
byte_stream_with_sps_extension.size(), &unit_stream));
|
||||
|
||||
const uint8_t kExpectedUnitStream[] = {
|
||||
0x00, 0x00, 0x00, 0x0A, 0x06, 0xFD, 0x78,
|
||||
0xA4, 0xC3, 0x82, 0x62, 0x11, 0x29, 0x77
|
||||
};
|
||||
EXPECT_EQ(std::vector<uint8_t>(std::begin(kExpectedUnitStream),
|
||||
std::end(kExpectedUnitStream)), unit_stream);
|
||||
|
||||
std::cout << std::endl << std::endl;
|
||||
std::vector<uint8_t> decoder_config;
|
||||
EXPECT_TRUE(converter.GetDecoderConfigurationRecord(&decoder_config));
|
||||
|
||||
const uint8_t kExpectedDecoderConfig[] = {
|
||||
0x01, 0x64, 0x00, 0x1E, 0xFF, 0xE1, 0x00, 0x1D,
|
||||
0x67, 0x64, 0x00, 0x1E, 0xAC, 0xD9, 0x40, 0xB4,
|
||||
0x2F, 0xF9, 0x7F, 0xF0, 0x00, 0x80, 0x00, 0x91,
|
||||
0x00, 0x00, 0x03, 0x03, 0xE9, 0x00, 0x00, 0xEA,
|
||||
0x60, 0x0F, 0x16, 0x2D, 0x96, 0x01, 0x00, 0x0A,
|
||||
0x68, 0xFE, 0xFD, 0xFC, 0xFB, 0x11, 0x12, 0x13,
|
||||
0x14, 0x15, 0xFD, 0xF8, 0xF8, 0x01, 0x6D, 0x33,
|
||||
0x01, 0x57, 0x78
|
||||
};
|
||||
EXPECT_EQ(std::vector<uint8_t>(std::begin(kExpectedDecoderConfig),
|
||||
std::end(kExpectedDecoderConfig)), decoder_config);
|
||||
}
|
||||
|
||||
} // namespace media
|
||||
} // namespace shaka
|
||||
|
|
|
|||
|
|
@ -242,6 +242,9 @@ bool NalUnitToByteStreamConverter::Initialize(
|
|||
buffer_writer.AppendArray(kNaluStartCode, arraysize(kNaluStartCode));
|
||||
AppendNalu(nalu, nalu_length_size_, !kEscapeData, &buffer_writer);
|
||||
found_pps = true;
|
||||
} else if (nalu.type() == Nalu::H264NaluType::H264_SPSExtension) {
|
||||
buffer_writer.AppendArray(kNaluStartCode, arraysize(kNaluStartCode));
|
||||
AppendNalu(nalu, nalu_length_size_, !kEscapeData, &buffer_writer);
|
||||
}
|
||||
}
|
||||
if (!found_sps || !found_pps) {
|
||||
|
|
@ -313,6 +316,8 @@ bool NalUnitToByteStreamConverter::ConvertUnitToByteStreamWithSubsamples(
|
|||
break;
|
||||
case Nalu::H264_SPS:
|
||||
FALLTHROUGH_INTENDED;
|
||||
case Nalu::H264_SPSExtension:
|
||||
FALLTHROUGH_INTENDED;
|
||||
case Nalu::H264_PPS: {
|
||||
// Also write this SPS/PPS if it is not the same as SPS/PPS in decoder
|
||||
// configuration, which is already written.
|
||||
|
|
|
|||
|
|
@ -201,6 +201,74 @@ TEST(NalUnitToByteStreamConverterTest, ConvertUnitToByteStream) {
|
|||
output);
|
||||
}
|
||||
|
||||
// Expect a valid AVCDecoderConfigurationRecord with SPSExtension to pass.
|
||||
TEST(NalUnitToByteStreamConverterTest, ConvertUnitToByteStreamWithSPSExtension) {
|
||||
NalUnitToByteStreamConverter converter;
|
||||
const uint8_t kDecoderConfigWithSpsExt[] = {
|
||||
0x01, // configuration version (must be 1)
|
||||
0x64, // AVCProfileIndication (100: sps special case)
|
||||
0x00, // profile_compatibility (bogus)
|
||||
0x00, // AVCLevelIndication (bogus)
|
||||
0xFF, // Length size minus 1 == 3
|
||||
0xE1, // 1 sps.
|
||||
0x00, 0x1D, // SPS length == 29
|
||||
// Some valid SPS data.
|
||||
0x67, 0x64, 0x00, 0x1E, 0xAC, 0xD9, 0x40, 0xB4,
|
||||
0x2F, 0xF9, 0x7F, 0xF0, 0x00, 0x80, 0x00, 0x91,
|
||||
0x00, 0x00, 0x03, 0x03, 0xE9, 0x00, 0x00, 0xEA,
|
||||
0x60, 0x0F, 0x16, 0x2D, 0x96,
|
||||
0x01, // 1 pps.
|
||||
0x00, 0x0A, // PPS length == 10
|
||||
0x68, 0xFE, 0xFD, 0xFC, 0xFB, 0x11, 0x12, 0x13, 0x14, 0x15,
|
||||
0xFC, // chroma_format == 0
|
||||
0xF9, // bit_depth_luma_minus8 == 1
|
||||
0xFF, // bit_depth_chroma_minus8 == 7
|
||||
0x01, // 1 SPS Extension
|
||||
0x00, 0x05, // SPSExtension length = 5
|
||||
0x6D, 0x33, 0x01, 0x57, 0x78
|
||||
};
|
||||
|
||||
// Only the type of the NAL units are checked.
|
||||
// This does not contain AUD, SPS, nor PPS.
|
||||
const uint8_t kUnitStreamLikeMediaSample[] = {
|
||||
0x00, 0x00, 0x00, 0x0A, // Size 10 NALU.
|
||||
0x06, // NAL unit type.
|
||||
0xFD, 0x78, 0xA4, 0xC3, 0x82, 0x62, 0x11, 0x29, 0x77
|
||||
};
|
||||
|
||||
const uint8_t kByteStreamWithSpsExtension[] = {
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
0x09, // AUD Type
|
||||
0xF0, // Primary pic type
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// Some SPS Data
|
||||
0x67, 0x64, 0x00, 0x1E, 0xAC, 0xD9, 0x40, 0xB4, 0x2F,
|
||||
0xF9, 0x7F, 0xF0, 0x00, 0x80, 0x00, 0x91, 0x00, 0x00,
|
||||
0x03, 0x03, 0xE9, 0x00, 0x00, 0xEA, 0x60, 0x0F, 0x16,
|
||||
0x2D, 0x96,
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// Some PPS Data
|
||||
0x68, 0xFE, 0xFD, 0xFC, 0xFB, 0x11, 0x12, 0x13, 0x14, 0x15,
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// Some SPS Extension data
|
||||
0x6D, 0x33, 0x01, 0x57, 0x78,
|
||||
0x00, 0x00, 0x00, 0x01, // Start code
|
||||
// The input NALU
|
||||
0x06, // NALU type
|
||||
0xFD, 0x78, 0xA4, 0xC3, 0x82, 0x62, 0x11, 0x29, 0x77,
|
||||
};
|
||||
EXPECT_TRUE(converter.Initialize(kDecoderConfigWithSpsExt,
|
||||
arraysize(kDecoderConfigWithSpsExt)));
|
||||
|
||||
std::vector<uint8_t> output;
|
||||
EXPECT_TRUE(converter.ConvertUnitToByteStream(
|
||||
kUnitStreamLikeMediaSample,
|
||||
arraysize(kUnitStreamLikeMediaSample), kIsKeyFrame, &output));
|
||||
EXPECT_EQ(std::vector<uint8_t>(std::begin(kByteStreamWithSpsExtension),
|
||||
std::end(kByteStreamWithSpsExtension)),
|
||||
output);
|
||||
}
|
||||
|
||||
// Verify that if it is not a key frame then SPS and PPS from decoder
|
||||
// configuration is not used.
|
||||
TEST(NalUnitToByteStreamConverterTest, NonKeyFrameSample) {
|
||||
|
|
|
|||
|
|
@ -35,8 +35,7 @@ bool H264VideoSliceHeaderParser::Initialize(
|
|||
const Nalu& nalu = config.nalu(i);
|
||||
if (nalu.type() == Nalu::H264_SPS) {
|
||||
RCHECK(parser_.ParseSps(nalu, &id) == H264Parser::kOk);
|
||||
} else {
|
||||
DCHECK_EQ(Nalu::H264_PPS, nalu.type());
|
||||
} else if (nalu.type() == Nalu::H264_PPS) {
|
||||
RCHECK(parser_.ParsePps(nalu, &id) == H264Parser::kOk);
|
||||
}
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in New Issue