Add RTP header extension for HDR metadata

Bug: webrtc:8651
Change-Id: I1c956eaac1532ac0d3820084edb4054a4cc9c68d
Reviewed-on: https://webrtc-review.googlesource.com/c/109924
Commit-Queue: Johannes Kron <kron@webrtc.org>
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Reviewed-by: Åsa Persson <asapersson@webrtc.org>
Reviewed-by: Alex Loiko <aleloi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25578}
This commit is contained in:
Johannes Kron 2018-11-09 11:12:36 +01:00 committed by Commit Bot
parent ee45f900c4
commit ad1d9f0d78
9 changed files with 208 additions and 0 deletions

View File

@ -15,7 +15,9 @@
#include <stdint.h>
#include <string.h>
#include "absl/types/optional.h"
#include "api/array_view.h"
#include "api/video/hdr_metadata.h"
#include "api/video/video_content_type.h"
#include "api/video/video_frame_marking.h"
#include "api/video/video_rotation.h"
@ -126,6 +128,8 @@ struct RTPHeaderExtension {
// For identifying the media section used to interpret this RTP packet. See
// https://tools.ietf.org/html/draft-ietf-mmusic-sdp-bundle-negotiation-38
Mid mid;
absl::optional<HdrMetadata> hdr_metadata;
};
struct RTPHeader {

View File

@ -112,6 +112,7 @@ enum RTPExtensionType : int {
kRtpExtensionRepairedRtpStreamId,
kRtpExtensionMid,
kRtpExtensionGenericFrameDescriptor,
kRtpExtensionHdrMetadata,
kRtpExtensionNumberOfExtensions // Must be the last entity in the enum.
};

View File

@ -43,6 +43,7 @@ constexpr ExtensionInfo kExtensions[] = {
CreateExtensionInfo<RepairedRtpStreamId>(),
CreateExtensionInfo<RtpMid>(),
CreateExtensionInfo<RtpGenericFrameDescriptorExtension>(),
CreateExtensionInfo<HdrMetadataExtension>(),
};
// Because of kRtpExtensionNone, NumberOfExtension is 1 bigger than the actual

View File

@ -11,6 +11,7 @@
#include "modules/rtp_rtcp/source/rtp_header_extensions.h"
#include <string.h>
#include <cmath>
#include "modules/rtp_rtcp/include/rtp_cvo.h"
#include "modules/rtp_rtcp/source/byte_io.h"
@ -433,6 +434,135 @@ bool FrameMarkingExtension::Write(rtc::ArrayView<uint8_t> data,
return true;
}
// HDR Metadata.
//
// RTP header extension to carry HDR metadata.
// Float values are upscaled by a static factor and transmitted as integers.
//
// 0 1 2 3
// 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | ID | length | luminance_max |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | | luminance_min |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | mastering_metadata.primary_r.x and .y |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | mastering_metadata.primary_g.x and .y |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | mastering_metadata.primary_b.x and .y |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | mastering_metadata.white.x and .y |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | max_content_light_level |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// | max_frame_average_light_level |
// +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
constexpr RTPExtensionType HdrMetadataExtension::kId;
constexpr uint8_t HdrMetadataExtension::kValueSizeBytes;
constexpr const char HdrMetadataExtension::kUri[];
bool HdrMetadataExtension::Parse(rtc::ArrayView<const uint8_t> data,
HdrMetadata* hdr_metadata) {
RTC_DCHECK(hdr_metadata);
if (data.size() != kValueSizeBytes)
return false;
size_t offset = 0;
offset += ParseLuminance(data.data() + offset,
&hdr_metadata->mastering_metadata.luminance_max,
kLuminanceMaxDenominator);
offset += ParseLuminance(data.data() + offset,
&hdr_metadata->mastering_metadata.luminance_min,
kLuminanceMinDenominator);
offset += ParseChromaticity(data.data() + offset,
&hdr_metadata->mastering_metadata.primary_r);
offset += ParseChromaticity(data.data() + offset,
&hdr_metadata->mastering_metadata.primary_g);
offset += ParseChromaticity(data.data() + offset,
&hdr_metadata->mastering_metadata.primary_b);
offset += ParseChromaticity(data.data() + offset,
&hdr_metadata->mastering_metadata.white_point);
// TODO(kron): Do we need 32 bit here or is it enough with 16 bits?
// Also, what resolution is needed?
hdr_metadata->max_content_light_level =
ByteReader<uint32_t>::ReadBigEndian(data.data() + offset);
offset += 4;
hdr_metadata->max_frame_average_light_level =
ByteReader<uint32_t>::ReadBigEndian(data.data() + offset);
RTC_DCHECK_EQ(kValueSizeBytes, offset + 4);
return true;
}
bool HdrMetadataExtension::Write(rtc::ArrayView<uint8_t> data,
const HdrMetadata& hdr_metadata) {
RTC_DCHECK_EQ(data.size(), kValueSizeBytes);
size_t offset = 0;
offset += WriteLuminance(data.data() + offset,
hdr_metadata.mastering_metadata.luminance_max,
kLuminanceMaxDenominator);
offset += WriteLuminance(data.data() + offset,
hdr_metadata.mastering_metadata.luminance_min,
kLuminanceMinDenominator);
offset += WriteChromaticity(data.data() + offset,
hdr_metadata.mastering_metadata.primary_r);
offset += WriteChromaticity(data.data() + offset,
hdr_metadata.mastering_metadata.primary_g);
offset += WriteChromaticity(data.data() + offset,
hdr_metadata.mastering_metadata.primary_b);
offset += WriteChromaticity(data.data() + offset,
hdr_metadata.mastering_metadata.white_point);
// TODO(kron): Do we need 32 bit here or is it enough with 16 bits?
// Also, what resolution is needed?
ByteWriter<uint32_t>::WriteBigEndian(data.data() + offset,
hdr_metadata.max_content_light_level);
offset += 4;
ByteWriter<uint32_t>::WriteBigEndian(
data.data() + offset, hdr_metadata.max_frame_average_light_level);
RTC_DCHECK_EQ(kValueSizeBytes, offset + 4);
return true;
}
size_t HdrMetadataExtension::ParseChromaticity(
const uint8_t* data,
HdrMasteringMetadata::Chromaticity* p) {
uint16_t chromaticity_x_scaled = ByteReader<uint16_t>::ReadBigEndian(data);
uint16_t chromaticity_y_scaled =
ByteReader<uint16_t>::ReadBigEndian(data + 2);
p->x = static_cast<float>(chromaticity_x_scaled) / kChromaticityDenominator;
p->y = static_cast<float>(chromaticity_y_scaled) / kChromaticityDenominator;
return 4; // Return number of bytes read.
}
size_t HdrMetadataExtension::ParseLuminance(const uint8_t* data,
float* f,
int denominator) {
uint32_t luminance_scaled = ByteReader<uint32_t, 3>::ReadBigEndian(data);
*f = static_cast<float>(luminance_scaled) / denominator;
return 3; // Return number of bytes read.
}
size_t HdrMetadataExtension::WriteChromaticity(
uint8_t* data,
const HdrMasteringMetadata::Chromaticity& p) {
RTC_DCHECK_GE(p.x, 0.0f);
RTC_DCHECK_GE(p.y, 0.0f);
ByteWriter<uint16_t>::WriteBigEndian(
data, std::round(p.x * kChromaticityDenominator));
ByteWriter<uint16_t>::WriteBigEndian(
data + 2, std::round(p.y * kChromaticityDenominator));
return 4; // Return number of bytes written.
}
size_t HdrMetadataExtension::WriteLuminance(uint8_t* data,
float f,
int denominator) {
RTC_DCHECK_GE(f, 0.0f);
ByteWriter<uint32_t, 3>::WriteBigEndian(data, std::round(f * denominator));
return 3; // Return number of bytes written.
}
bool BaseRtpStringExtension::Parse(rtc::ArrayView<const uint8_t> data,
StringRtpHeaderExtension* str) {
if (data.empty() || data[0] == 0) // Valid string extension can't be empty.

View File

@ -16,6 +16,7 @@
#include "api/array_view.h"
#include "api/rtp_headers.h"
#include "api/video/hdr_metadata.h"
#include "api/video/video_content_type.h"
#include "api/video/video_frame_marking.h"
#include "api/video/video_rotation.h"
@ -181,6 +182,32 @@ class FrameMarkingExtension {
static bool IsScalable(uint8_t temporal_id, uint8_t layer_id);
};
class HdrMetadataExtension {
public:
using value_type = HdrMetadata;
static constexpr RTPExtensionType kId = kRtpExtensionHdrMetadata;
static constexpr uint8_t kValueSizeBytes = 30;
// TODO(webrtc:8651): Change to a valid uri.
static constexpr const char kUri[] = "rtp-hdr-metadata-uri-placeholder";
static bool Parse(rtc::ArrayView<const uint8_t> data,
HdrMetadata* hdr_metadata);
static size_t ValueSize(const HdrMetadata&) { return kValueSizeBytes; }
static bool Write(rtc::ArrayView<uint8_t> data,
const HdrMetadata& hdr_metadata);
private:
static constexpr int kChromaticityDenominator = 10000; // 0.0001 resolution.
static constexpr int kLuminanceMaxDenominator = 100; // 0.01 resolution.
static constexpr int kLuminanceMinDenominator = 10000; // 0.0001 resolution.
static size_t ParseChromaticity(const uint8_t* data,
HdrMasteringMetadata::Chromaticity* p);
static size_t ParseLuminance(const uint8_t* data, float* f, int denominator);
static size_t WriteChromaticity(uint8_t* data,
const HdrMasteringMetadata::Chromaticity& p);
static size_t WriteLuminance(uint8_t* data, float f, int denominator);
};
// Base extension class for RTP header extensions which are strings.
// Subclasses must defined kId and kUri static constexpr members.
class BaseRtpStringExtension {

View File

@ -69,6 +69,7 @@ void RtpPacketReceived::GetHeader(RTPHeader* header) const {
GetExtension<RepairedRtpStreamId>(&header->extension.repaired_stream_id);
GetExtension<RtpMid>(&header->extension.mid);
GetExtension<PlayoutDelayLimits>(&header->extension.playout_delay);
header->extension.hdr_metadata = GetExtension<HdrMetadataExtension>();
}
} // namespace webrtc

View File

@ -185,6 +185,24 @@ constexpr uint8_t kPacketWithLegacyTimingExtension[] = {
0x04, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00};
// clang-format on
HdrMetadata CreateTestHdrMetadata() {
// Random but reasonable HDR metadata.
HdrMetadata hdr_metadata;
hdr_metadata.mastering_metadata.luminance_max = 2000.0;
hdr_metadata.mastering_metadata.luminance_min = 2.0001;
hdr_metadata.mastering_metadata.primary_r.x = 0.3003;
hdr_metadata.mastering_metadata.primary_r.y = 0.4004;
hdr_metadata.mastering_metadata.primary_g.x = 0.3201;
hdr_metadata.mastering_metadata.primary_g.y = 0.4604;
hdr_metadata.mastering_metadata.primary_b.x = 0.3409;
hdr_metadata.mastering_metadata.primary_b.y = 0.4907;
hdr_metadata.mastering_metadata.white_point.x = 0.4103;
hdr_metadata.mastering_metadata.white_point.y = 0.4806;
hdr_metadata.max_content_light_level = 2345;
hdr_metadata.max_frame_average_light_level = 1789;
return hdr_metadata;
}
} // namespace
TEST(RtpPacketTest, CreateMinimum) {
@ -801,4 +819,21 @@ TEST(RtpPacketTest, ParseLegacyTimingFrameExtension) {
EXPECT_EQ(receivied_timing.flags, 0);
}
TEST(RtpPacketTest, CreateAndParseHdrMetadataExtension) {
// Create packet with extension.
RtpPacket::ExtensionManager extensions(/*extmap-allow-mixed=*/true);
extensions.Register<HdrMetadataExtension>(1);
RtpPacket packet(&extensions);
const HdrMetadata kHdrMetadata = CreateTestHdrMetadata();
EXPECT_TRUE(packet.SetExtension<HdrMetadataExtension>(kHdrMetadata));
packet.SetPayloadSize(42);
// Read packet with the extension.
RtpPacketReceived parsed(&extensions);
EXPECT_TRUE(parsed.Parse(packet.Buffer()));
HdrMetadata parsed_hdr_metadata;
EXPECT_TRUE(parsed.GetExtension<HdrMetadataExtension>(&parsed_hdr_metadata));
EXPECT_EQ(kHdrMetadata, parsed_hdr_metadata);
}
} // namespace webrtc

View File

@ -507,6 +507,10 @@ void RtpHeaderParser::ParseOneByteExtensionHeader(
RTC_LOG(WARNING)
<< "RtpGenericFrameDescriptor unsupported by rtp header parser.";
break;
case kRtpExtensionHdrMetadata:
RTC_LOG(WARNING)
<< "RtpExtensionHdrMetadata unsupported by rtp header parser.";
break;
case kRtpExtensionNone:
case kRtpExtensionNumberOfExtensions: {
RTC_NOTREACHED() << "Invalid extension type: " << type;

View File

@ -121,6 +121,11 @@ void FuzzOneInput(const uint8_t* data, size_t size) {
packet.GetExtension<RtpGenericFrameDescriptorExtension>(&descriptor);
break;
}
case kRtpExtensionHdrMetadata: {
HdrMetadata hdr_metadata;
packet.GetExtension<HdrMetadataExtension>(&hdr_metadata);
break;
}
}
}
}