zqz/platform-external-webrtc
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add h.264 AVC SPS parsing for resolution.

Browse Source 
BUG=
R=stefan@webrtc.org

Review URL: https://webrtc-codereview.appspot.com/48129004

Cr-Commit-Position: refs/heads/master@{#9073}
This commit is contained in:
Noah Richards
2015-04-23 16:45:56 -07:00
parent 9728241e6a
commit 5ea8eff55e
8 changed files with 389 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
webrtc/modules/modules.gyp
View File

@ -223,6 +223,7 @@
'rtp_rtcp/source/fec_receiver_unittest.cc',
'rtp_rtcp/source/fec_test_helper.cc',
'rtp_rtcp/source/fec_test_helper.h',
'rtp_rtcp/source/h264_sps_parser_unittest.cc',
'rtp_rtcp/source/nack_rtx_unittest.cc',
'rtp_rtcp/source/producer_fec_unittest.cc',
'rtp_rtcp/source/receive_statistics_unittest.cc',

2
webrtc/modules/rtp_rtcp/BUILD.gn
View File

@ -67,6 +67,8 @@ source_set("rtp_rtcp") {
"source/forward_error_correction.h",
"source/forward_error_correction_internal.cc",
"source/forward_error_correction_internal.h",
"source/h264_sps_parser.cc",
"source/h264_sps_parser.h",
"source/producer_fec.cc",
"source/producer_fec.h",
"source/rtp_packet_history.cc",

2
webrtc/modules/rtp_rtcp/rtp_rtcp.gypi
View File

@ -74,6 +74,8 @@
'source/forward_error_correction.h',
'source/forward_error_correction_internal.cc',
'source/forward_error_correction_internal.h',
'source/h264_sps_parser.cc',
'source/h264_sps_parser.h',
'source/producer_fec.cc',
'source/producer_fec.h',
'source/rtp_packet_history.cc',

226
webrtc/modules/rtp_rtcp/source/h264_sps_parser.cc Normal file
View File

@ -0,0 +1,226 @@
/*
* Copyright (c) 2015 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "webrtc/modules/rtp_rtcp/source/h264_sps_parser.h"
#include "webrtc/base/bytebuffer.h"
#include "webrtc/base/bitbuffer.h"
#include "webrtc/system_wrappers/interface/logging.h"
#define RETURN_FALSE_ON_FAIL(x) \
if (!(x)) { \
return false; \
}
namespace webrtc {
H264SpsParser::H264SpsParser(const uint8* sps, size_t byte_length)
: sps_(sps), byte_length_(byte_length), width_(), height_() {
}
bool H264SpsParser::Parse() {
// General note: this is based off the 02/2014 version of the H.264 standard.
// You can find it on this page:
// http://www.itu.int/rec/T-REC-H.264
const char* sps_bytes = reinterpret_cast<const char*>(sps_);
// First, parse out rbsp, which is basically the source buffer minus emulation
// bytes (0x03). RBSP is defined in section 7.3.1 of the H.264 standard.
rtc::ByteBuffer rbsp_buffer;
for (size_t i = 0; i < byte_length_;) {
if (i < byte_length_ - 3 && sps_[i + 3] == 3) {
// Two rbsp bytes + the emulation byte.
rbsp_buffer.WriteBytes(sps_bytes + i, 2);
i += 3;
} else {
// Single rbsp byte.
rbsp_buffer.WriteBytes(sps_bytes + i, 1);
i++;
}
}
// Now, we need to use a bit buffer to parse through the actual AVC SPS
// format. See Section 7.3.2.1.1 ("Sequence parameter set data syntax") of the
// H.264 standard for a complete description.
// Since we only care about resolution, we ignore the majority of fields, but
// we still have to actively parse through a lot of the data, since many of
// the fields have variable size.
// We're particularly interested in:
// chroma_format_idc -> affects crop units
// pic_{width,height}_* -> resolution of the frame in macroblocks (16x16).
// frame_crop_*_offset -> crop information
rtc::BitBuffer parser(reinterpret_cast<const uint8*>(rbsp_buffer.Data()),
rbsp_buffer.Length());
// The golomb values we have to read, not just consume.
uint32 golomb_ignored;
// separate_colour_plane_flag is optional (assumed 0), but has implications
// about the ChromaArrayType, which modifies how we treat crop coordinates.
uint32 separate_colour_plane_flag = 0;
// chroma_format_idc will be ChromaArrayType if separate_colour_plane_flag is
// 0. It defaults to 1, when not specified.
uint32 chroma_format_idc = 1;
// profile_idc: u(8). We need it to determine if we need to read/skip chroma
// formats.
uint8 profile_idc;
RETURN_FALSE_ON_FAIL(parser.ReadUInt8(&profile_idc));
// constraint_set0_flag through constraint_set5_flag + reserved_zero_2bits
// 1 bit each for the flags + 2 bits = 8 bits = 1 byte.
RETURN_FALSE_ON_FAIL(parser.ConsumeBytes(1));
// level_idc: u(8)
RETURN_FALSE_ON_FAIL(parser.ConsumeBytes(1));
// seq_parameter_set_id: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
// See if profile_idc has chroma format information.
if (profile_idc == 100 || profile_idc == 110 || profile_idc == 122 ||
profile_idc == 244 || profile_idc == 44 || profile_idc == 83 ||
profile_idc == 86 || profile_idc == 118 || profile_idc == 128 ||
profile_idc == 138 || profile_idc == 139 || profile_idc == 134) {
// chroma_format_idc: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&chroma_format_idc));
if (chroma_format_idc == 3) {
// separate_colour_plane_flag: u(1)
RETURN_FALSE_ON_FAIL(parser.ReadBits(&separate_colour_plane_flag, 1));
}
// bit_depth_luma_minus8: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
// bit_depth_chroma_minus8: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
// qpprime_y_zero_transform_bypass_flag: u(1)
RETURN_FALSE_ON_FAIL(parser.ConsumeBits(1));
// seq_scaling_matrix_present_flag: u(1)
uint32 seq_scaling_matrix_present_flag;
RETURN_FALSE_ON_FAIL(parser.ReadBits(&seq_scaling_matrix_present_flag, 1));
if (seq_scaling_matrix_present_flag) {
// seq_scaling_list_present_flags. Either 8 or 12, depending on
// chroma_format_idc.
uint32 seq_scaling_list_present_flags;
if (chroma_format_idc != 3) {
RETURN_FALSE_ON_FAIL(
parser.ReadBits(&seq_scaling_list_present_flags, 8));
} else {
RETURN_FALSE_ON_FAIL(
parser.ReadBits(&seq_scaling_list_present_flags, 12));
}
// We don't support reading the sequence scaling list, and we don't really
// see/use them in practice, so we'll just reject the full sps if we see
// any provided.
if (seq_scaling_list_present_flags > 0) {
LOG(LS_WARNING) << "SPS contains scaling lists, which are unsupported.";
return false;
}
}
}
// log2_max_frame_num_minus4: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
// pic_order_cnt_type: ue(v)
uint32 pic_order_cnt_type;
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&pic_order_cnt_type));
if (pic_order_cnt_type == 0) {
// log2_max_pic_order_cnt_lsb_minus4: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
} else if (pic_order_cnt_type == 1) {
// delta_pic_order_always_zero_flag: u(1)
RETURN_FALSE_ON_FAIL(parser.ConsumeBits(1));
// offset_for_non_ref_pic: se(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
// offset_for_top_to_bottom_field: se(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
// num_ref_frames_in_pic_order_cnt_cycle: ue(v)
uint32 num_ref_frames_in_pic_order_cnt_cycle;
RETURN_FALSE_ON_FAIL(
parser.ReadExponentialGolomb(&num_ref_frames_in_pic_order_cnt_cycle));
for (size_t i = 0; i < num_ref_frames_in_pic_order_cnt_cycle; ++i) {
// offset_for_ref_frame[i]: se(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
}
}
// max_num_ref_frames: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&golomb_ignored));
// gaps_in_frame_num_value_allowed_flag: u(1)
RETURN_FALSE_ON_FAIL(parser.ConsumeBits(1));
//
// IMPORTANT ONES! Now we're getting to resolution. First we read the pic
// width/height in macroblocks (16x16), which gives us the base resolution,
// and then we continue on until we hit the frame crop offsets, which are used
// to signify resolutions that aren't multiples of 16.
//
// pic_width_in_mbs_minus1: ue(v)
uint32 pic_width_in_mbs_minus1;
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&pic_width_in_mbs_minus1));
// pic_height_in_map_units_minus1: ue(v)
uint32 pic_height_in_map_units_minus1;
RETURN_FALSE_ON_FAIL(
parser.ReadExponentialGolomb(&pic_height_in_map_units_minus1));
// frame_mbs_only_flag: u(1)
uint32 frame_mbs_only_flag;
RETURN_FALSE_ON_FAIL(parser.ReadBits(&frame_mbs_only_flag, 1));
if (!frame_mbs_only_flag) {
// mb_adaptive_frame_field_flag: u(1)
RETURN_FALSE_ON_FAIL(parser.ConsumeBits(1));
}
// direct_8x8_inference_flag: u(1)
RETURN_FALSE_ON_FAIL(parser.ConsumeBits(1));
//
// MORE IMPORTANT ONES! Now we're at the frame crop information.
//
// frame_cropping_flag: u(1)
uint32 frame_cropping_flag;
uint32 frame_crop_left_offset = 0;
uint32 frame_crop_right_offset = 0;
uint32 frame_crop_top_offset = 0;
uint32 frame_crop_bottom_offset = 0;
RETURN_FALSE_ON_FAIL(parser.ReadBits(&frame_cropping_flag, 1));
if (frame_cropping_flag) {
// frame_crop_{left, right, top, bottom}_offset: ue(v)
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&frame_crop_left_offset));
RETURN_FALSE_ON_FAIL(
parser.ReadExponentialGolomb(&frame_crop_right_offset));
RETURN_FALSE_ON_FAIL(parser.ReadExponentialGolomb(&frame_crop_top_offset));
RETURN_FALSE_ON_FAIL(
parser.ReadExponentialGolomb(&frame_crop_bottom_offset));
}
// Far enough! We don't use the rest of the SPS.
// Start with the resolution determined by the pic_width/pic_height fields.
int width = 16 * (pic_width_in_mbs_minus1 + 1);
int height =
16 * (2 - frame_mbs_only_flag) * (pic_height_in_map_units_minus1 + 1);
// Figure out the crop units in pixels. That's based on the chroma format's
// sampling, which is indicated by chroma_format_idc.
if (separate_colour_plane_flag || chroma_format_idc == 0) {
frame_crop_bottom_offset *= (2 - frame_mbs_only_flag);
frame_crop_top_offset *= (2 - frame_mbs_only_flag);
} else if (!separate_colour_plane_flag && chroma_format_idc > 0) {
// Width multipliers for formats 1 (4:2:0) and 2 (4:2:2).
if (chroma_format_idc == 1 || chroma_format_idc == 2) {
frame_crop_left_offset *= 2;
frame_crop_right_offset *= 2;
}
// Height multipliers for format 1 (4:2:0).
if (chroma_format_idc == 1) {
frame_crop_top_offset *= 2;
frame_crop_bottom_offset *= 2;
}
}
// Subtract the crop for each dimension.
width -= (frame_crop_left_offset + frame_crop_right_offset);
height -= (frame_crop_top_offset + frame_crop_bottom_offset);
width_ = width;
height_ = height;
return true;
}
} // namespace webrtc

37
webrtc/modules/rtp_rtcp/source/h264_sps_parser.h Normal file
View File

@ -0,0 +1,37 @@
/*
* Copyright (c) 2015 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef WEBRTC_MODULES_RTP_RTCP_SOURCE_H264_SPS_PARSER_H_
#define WEBRTC_MODULES_RTP_RTCP_SOURCE_H264_SPS_PARSER_H_
#include "webrtc/base/common.h"
namespace webrtc {
// A class for parsing out sequence parameter set (SPS) data from an H264 NALU.
// Currently, only resolution is read without being ignored.
class H264SpsParser {
public:
H264SpsParser(const uint8* sps, size_t byte_length);
// Parses the SPS to completion. Returns true if the SPS was parsed correctly.
bool Parse();
uint16 width() { return width_; }
uint16 height() { return height_; }
private:
const uint8* const sps_;
const size_t byte_length_;
uint16 width_;
uint16 height_;
};
} // namespace webrtc
#endif // WEBRTC_MODULES_RTP_RTCP_SOURCE_H264_SPS_PARSER_H_

68
webrtc/modules/rtp_rtcp/source/h264_sps_parser_unittest.cc Normal file
View File

@ -0,0 +1,68 @@
/*
* Copyright (c) 2015 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "webrtc/modules/rtp_rtcp/source/h264_sps_parser.h"
#include "testing/gtest/include/gtest/gtest.h"
namespace webrtc {
// Example SPS can be generated with ffmpeg. Here's an example set of commands,
// runnable on OS X:
// 1) Generate a video, from the camera:
// ffmpeg -f avfoundation -i "0" -video_size 640x360 camera.mov
//
// 2) Scale the video to the desired size:
// ffmpeg -i camera.mov -vf scale=640x360 scaled.mov
//
// 3) Get just the H.264 bitstream in AnnexB:
// ffmpeg -i scaled.mov -vcodec copy -vbsf h264_mp4toannexb -an out.h264
//
// 4) Open out.h264 and find the SPS, generally everything between the first
// two start codes (0 0 0 1 or 0 0 1). The first byte should be 0x67,
// which should be stripped out before being passed to the parser.
TEST(H264SpsParserTest, TestSampleSPSHdLandscape) {
// SPS for a 1280x720 camera capture from ffmpeg on osx. Contains
// emulation bytes but no cropping.
const uint8 buffer[] = {0x7A, 0x00, 0x1F, 0xBC, 0xD9, 0x40, 0x50, 0x05,
0xBA, 0x10, 0x00, 0x00, 0x03, 0x00, 0xC0, 0x00,
0x00, 0x2A, 0xE0, 0xF1, 0x83, 0x19, 0x60};
H264SpsParser parser = H264SpsParser(buffer, ARRAY_SIZE(buffer));
EXPECT_TRUE(parser.Parse());
EXPECT_EQ(1280u, parser.width());
EXPECT_EQ(720u, parser.height());
}
TEST(H264SpsParserTest, TestSampleSPSVgaLandscape) {
// SPS for a 640x360 camera capture from ffmpeg on osx. Contains emulation
// bytes and cropping (360 isn't divisible by 16).
const uint8 buffer[] = {0x7A, 0x00, 0x1E, 0xBC, 0xD9, 0x40, 0xA0, 0x2F,
0xF8, 0x98, 0x40, 0x00, 0x00, 0x03, 0x01, 0x80,
0x00, 0x00, 0x56, 0x83, 0xC5, 0x8B, 0x65, 0x80};
H264SpsParser parser = H264SpsParser(buffer, ARRAY_SIZE(buffer));
EXPECT_TRUE(parser.Parse());
EXPECT_EQ(640u, parser.width());
EXPECT_EQ(360u, parser.height());
}
TEST(H264SpsParserTest, TestSampleSPSWeirdResolution) {
// SPS for a 200x400 camera capture from ffmpeg on osx. Horizontal and
// veritcal crop (neither dimension is divisible by 16).
const uint8 buffer[] = {0x7A, 0x00, 0x0D, 0xBC, 0xD9, 0x43, 0x43, 0x3E,
0x5E, 0x10, 0x00, 0x00, 0x03, 0x00, 0x60, 0x00,
0x00, 0x15, 0xA0, 0xF1, 0x42, 0x99, 0x60};
H264SpsParser parser = H264SpsParser(buffer, ARRAY_SIZE(buffer));
EXPECT_TRUE(parser.Parse());
EXPECT_EQ(200u, parser.width());
EXPECT_EQ(400u, parser.height());
}
} // namespace webrtc

17
webrtc/modules/rtp_rtcp/source/rtp_format_h264.cc
View File

@ -12,6 +12,7 @@
#include "webrtc/modules/interface/module_common_types.h"
#include "webrtc/modules/rtp_rtcp/source/byte_io.h"
#include "webrtc/modules/rtp_rtcp/source/h264_sps_parser.h"
#include "webrtc/modules/rtp_rtcp/source/rtp_format_h264.h"
namespace webrtc {
@ -30,6 +31,7 @@ enum Nalu {
static const size_t kNalHeaderSize = 1;
static const size_t kFuAHeaderSize = 2;
static const size_t kLengthFieldSize = 2;
static const size_t kStapAHeaderSize = kNalHeaderSize + kLengthFieldSize;
// Bit masks for FU (A and B) indicators.
enum NalDefs { kFBit = 0x80, kNriMask = 0x60, kTypeMask = 0x1F };
@ -47,15 +49,28 @@ void ParseSingleNalu(RtpDepacketizer::ParsedPayload* parsed_payload,
RTPVideoHeaderH264* h264_header =
&parsed_payload->type.Video.codecHeader.H264;
const uint8_t* nalu_start = payload_data + kNalHeaderSize;
size_t nalu_length = payload_data_length - kNalHeaderSize;
uint8_t nal_type = payload_data[0] & kTypeMask;
if (nal_type == kStapA) {
nal_type = payload_data[3] & kTypeMask;
// Skip the StapA header (StapA nal type + length).
nal_type = payload_data[kStapAHeaderSize] & kTypeMask;
nalu_start += kStapAHeaderSize;
nalu_length -= kStapAHeaderSize;
h264_header->packetization_type = kH264StapA;
} else {
h264_header->packetization_type = kH264SingleNalu;
}
h264_header->nalu_type = nal_type;
// We can read resolution out of sps packets.
if (nal_type == kSps) {
H264SpsParser parser(nalu_start, nalu_length);
if (parser.Parse()) {
parsed_payload->type.Video.width = parser.width();
parsed_payload->type.Video.height = parser.height();
}
}
switch (nal_type) {
case kSps:
case kPps:

37
webrtc/modules/rtp_rtcp/source/rtp_format_h264_unittest.cc
View File

@ -411,6 +411,23 @@ TEST_F(RtpDepacketizerH264Test, TestSingleNalu) {
EXPECT_EQ(kIdr, payload.type.Video.codecHeader.H264.nalu_type);
}
TEST_F(RtpDepacketizerH264Test, TestSingleNaluSpsWithResolution) {
uint8_t packet[] = {kSps, 0x7A, 0x00, 0x1F, 0xBC, 0xD9, 0x40, 0x50,
0x05, 0xBA, 0x10, 0x00, 0x00, 0x03, 0x00, 0xC0,
0x00, 0x00, 0x2A, 0xE0, 0xF1, 0x83, 0x19, 0x60};
RtpDepacketizer::ParsedPayload payload;
ASSERT_TRUE(depacketizer_->Parse(&payload, packet, sizeof(packet)));
ExpectPacket(&payload, packet, sizeof(packet));
EXPECT_EQ(kVideoFrameKey, payload.frame_type);
EXPECT_EQ(kRtpVideoH264, payload.type.Video.codec);
EXPECT_TRUE(payload.type.Video.isFirstPacket);
EXPECT_EQ(kH264SingleNalu,
payload.type.Video.codecHeader.H264.packetization_type);
EXPECT_EQ(1280u, payload.type.Video.width);
EXPECT_EQ(720u, payload.type.Video.height);
}
TEST_F(RtpDepacketizerH264Test, TestStapAKey) {
uint8_t packet[16] = {kStapA, // F=0, NRI=0, Type=24.
// Length, nal header, payload.
@ -429,6 +446,26 @@ TEST_F(RtpDepacketizerH264Test, TestStapAKey) {
EXPECT_EQ(kSps, payload.type.Video.codecHeader.H264.nalu_type);
}
TEST_F(RtpDepacketizerH264Test, TestStapANaluSpsWithResolution) {
uint8_t packet[] = {kStapA, // F=0, NRI=0, Type=24.
// Length (2 bytes), nal header, payload.
0, 24, kSps, 0x7A, 0x00, 0x1F, 0xBC, 0xD9,
0x40, 0x50, 0x05, 0xBA, 0x10, 0x00, 0x00, 0x03,
0x00, 0xC0, 0x00, 0x00, 0x2A, 0xE0, 0xF1, 0x83,
0x19, 0x60, 0, 0x03, kIdr, 0xFF, 0x00, 0,
0x04, kIdr, 0xFF, 0x00, 0x11};
RtpDepacketizer::ParsedPayload payload;
ASSERT_TRUE(depacketizer_->Parse(&payload, packet, sizeof(packet)));
ExpectPacket(&payload, packet, sizeof(packet));
EXPECT_EQ(kVideoFrameKey, payload.frame_type);
EXPECT_EQ(kRtpVideoH264, payload.type.Video.codec);
EXPECT_TRUE(payload.type.Video.isFirstPacket);
EXPECT_EQ(kH264StapA, payload.type.Video.codecHeader.H264.packetization_type);
EXPECT_EQ(1280u, payload.type.Video.width);
EXPECT_EQ(720u, payload.type.Video.height);
}
TEST_F(RtpDepacketizerH264Test, TestStapADelta) {
uint8_t packet[16] = {kStapA, // F=0, NRI=0, Type=24.
// Length, nal header, payload.