avcodec: Pass HDR10+ metadata to packet side data in VP9 encoder

HDR10+ metadata is stored in the bit stream for HEVC. The story is different for VP9 and cannot store the metadata in the bit stream. HDR10+ should be passed to packet side data an stored in the container (mkv) for VP9. This CL is taking HDR10+ from AVFrame side data in libvpxenc and is passing it to the AVPacket side data.

Mohammad Izadi committed 5 years ago

Commit aca923b1 parent 947122f6 files · +105 −2Browse files

Unified · Split

6 files

doc/APIchanges +2 −0
libavcodec/avpacket.c +1 −0
libavcodec/decode.c +1 −0
libavcodec/libvpxenc.c +91 −0
libavcodec/packet.h +8 −0
libavcodec/version.h +2 −2

doc/APIchanges+2−0

@@ -13,6 +13,8 @@ libavutil: 2021-04-27 · 12 unmodified lines
13		13
14		14
15	API changes, most recent first:	15	API changes, most recent first:
		16	2021-06-17 - xxxxxxxxxx - lavc 59.2.100 - packet.h
		17	Add AV_PKT_DATA_DYNAMIC_HDR10_PLUS
16		18
17	2021-06-09 - xxxxxxxxxx - lavf 59.3.100 - avformat.h	19	2021-06-09 - xxxxxxxxxx - lavf 59.3.100 - avformat.h
18	Add pts_wrap_bits to AVStream	20	Add pts_wrap_bits to AVStream

libavcodec/avpacket.c+1−0

@@ -289,6 +289,7 @@ const char *av_packet_side_data_name(enum AVPacketSideDataType type) · 288 unmodified lines
289	case AV_PKT_DATA_ICC_PROFILE: return "ICC Profile";	289	case AV_PKT_DATA_ICC_PROFILE: return "ICC Profile";
290	case AV_PKT_DATA_DOVI_CONF: return "DOVI configuration record";	290	case AV_PKT_DATA_DOVI_CONF: return "DOVI configuration record";
291	case AV_PKT_DATA_S12M_TIMECODE: return "SMPTE ST 12-1:2014 timecode";	291	case AV_PKT_DATA_S12M_TIMECODE: return "SMPTE ST 12-1:2014 timecode";
		292	case AV_PKT_DATA_DYNAMIC_HDR10_PLUS: return "HDR10+ Dynamic Metadata (SMPTE 2094-40)";
292	}	293	}
293	return NULL;	294	return NULL;
294	}	295	}

libavcodec/decode.c+1−0

@@ -1497,6 +1497,7 @@ int ff_decode_frame_props(AVCodecContext avctx, AVFrame frame) · 1496 unmodified lines
1497	{ AV_PKT_DATA_A53_CC, AV_FRAME_DATA_A53_CC },	1497	{ AV_PKT_DATA_A53_CC, AV_FRAME_DATA_A53_CC },
1498	{ AV_PKT_DATA_ICC_PROFILE, AV_FRAME_DATA_ICC_PROFILE },	1498	{ AV_PKT_DATA_ICC_PROFILE, AV_FRAME_DATA_ICC_PROFILE },
1499	{ AV_PKT_DATA_S12M_TIMECODE, AV_FRAME_DATA_S12M_TIMECODE },	1499	{ AV_PKT_DATA_S12M_TIMECODE, AV_FRAME_DATA_S12M_TIMECODE },
		1500	{ AV_PKT_DATA_DYNAMIC_HDR10_PLUS, AV_FRAME_DATA_DYNAMIC_HDR_PLUS },
1500	};	1501	};
1501		1502
1502	if (!(avctx->codec->caps_internal & FF_CODEC_CAP_SETS_FRAME_PROPS)) {	1503	if (!(avctx->codec->caps_internal & FF_CODEC_CAP_SETS_FRAME_PROPS)) {

libavcodec/libvpxenc.c+91−0

@@ -64,6 +64,11 @@ struct FrameListData { · 63 unmodified lines
64	struct FrameListData *next;	64	struct FrameListData *next;
65	};	65	};
66		66
		67	typedef struct FrameHDR10Plus {
		68	int64_t pts;
		69	AVBufferRef *hdr10_plus;
		70	} FrameHDR10Plus;
		71
67	typedef struct VPxEncoderContext {	72	typedef struct VPxEncoderContext {
68	AVClass *class;	73	AVClass *class;
69	struct vpx_codec_ctx encoder;	74	struct vpx_codec_ctx encoder;
@@ -121,6 +126,8 @@ typedef struct VPxEncoderContext { · 51 unmodified lines
121	int tune_content;	126	int tune_content;
122	int corpus_complexity;	127	int corpus_complexity;
123	int tpl_model;	128	int tpl_model;
		129	int discard_hdr10_plus;
		130	AVFifoBuffer *hdr10_plus_fifo;
124	/**	131	/**
125	* If the driver does not support ROI then warn the first time we	132	* If the driver does not support ROI then warn the first time we
126	* encounter a frame with ROI side data.	133	* encounter a frame with ROI side data.
@@ -316,6 +323,51 @@ static av_cold void free_frame_list(struct FrameListData *list) · 189 unmodified lines
316	}	323	}
317	}	324	}
318		325
		326	static av_cold int add_hdr10_plus(AVFifoBuffer fifo, struct FrameHDR10Plus data)
		327	{
		328	int err = av_fifo_grow(fifo, sizeof(*data));
		329	if (err < 0)
		330	return err;
		331	av_fifo_generic_write(fifo, data, sizeof(*data), NULL);
		332	return 0;
		333	}
		334
		335	static av_cold void free_hdr10_plus_fifo(AVFifoBuffer **fifo)
		336	{
		337	FrameHDR10Plus frame_hdr10_plus;
		338	while (av_fifo_size(*fifo) >= sizeof(frame_hdr10_plus)) {
		339	av_fifo_generic_read(*fifo, &frame_hdr10_plus, sizeof(frame_hdr10_plus), NULL);
		340	av_buffer_unref(&frame_hdr10_plus.hdr10_plus);
		341	}
		342	av_fifo_freep(fifo);
		343	}
		344
		345	static int copy_hdr10_plus_to_pkt(AVFifoBuffer fifo, AVPacket pkt)
		346	{
		347	FrameHDR10Plus frame_hdr10_plus;
		348	uint8_t *data;
		349	if (!pkt)
		350	return 0;
		351	if (av_fifo_size(fifo) < sizeof(frame_hdr10_plus))
		352	return 0;
		353	av_fifo_generic_peek(fifo, &frame_hdr10_plus, sizeof(frame_hdr10_plus), NULL);
		354	if (!frame_hdr10_plus.hdr10_plus \|\| frame_hdr10_plus.pts != pkt->pts)
		355	return 0;
		356	av_fifo_generic_read(fifo, &frame_hdr10_plus, sizeof(frame_hdr10_plus), NULL);
		357	if (!frame_hdr10_plus.hdr10_plus)
		358	return 0;
		359
		360	data = av_packet_new_side_data(pkt, AV_PKT_DATA_DYNAMIC_HDR10_PLUS, frame_hdr10_plus.hdr10_plus->size);
		361	if (!data) {
		362	av_buffer_unref(&frame_hdr10_plus.hdr10_plus);
		363	return AVERROR(ENOMEM);
		364	}
		365
		366	memcpy(data, frame_hdr10_plus.hdr10_plus->data, frame_hdr10_plus.hdr10_plus->size);
		367	av_buffer_unref(&frame_hdr10_plus.hdr10_plus);
		368	return 0;
		369	}
		370
319	static av_cold int codecctl_int(AVCodecContext *avctx,	371	static av_cold int codecctl_int(AVCodecContext *avctx,
320	enum vp8e_enc_control_id id, int val)	372	enum vp8e_enc_control_id id, int val)
321	{	373	{
@@ -384,6 +436,8 @@ static av_cold int vpx_free(AVCodecContext *avctx) · 62 unmodified lines
384	av_freep(&ctx->twopass_stats.buf);	436	av_freep(&ctx->twopass_stats.buf);
385	av_freep(&avctx->stats_out);	437	av_freep(&avctx->stats_out);
386	free_frame_list(ctx->coded_frame_list);	438	free_frame_list(ctx->coded_frame_list);
		439	if (ctx->hdr10_plus_fifo)
		440	free_hdr10_plus_fifo(&ctx->hdr10_plus_fifo);
387	return 0;	441	return 0;
388	}	442	}
389		443
@@ -835,6 +889,7 @@ static av_cold int vpx_init(AVCodecContext *avctx, · 445 unmodified lines
835	#endif	889	#endif
836	AVDictionaryEntry* en = NULL;	890	AVDictionaryEntry* en = NULL;
837		891
		892	ctx->discard_hdr10_plus = 1;
838	av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());	893	av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
839	av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());	894	av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
840		895
@@ -851,6 +906,14 @@ static av_cold int vpx_init(AVCodecContext *avctx, · 10 unmodified lines
851	if (avctx->codec_id == AV_CODEC_ID_VP9) {	906	if (avctx->codec_id == AV_CODEC_ID_VP9) {
852	if (set_pix_fmt(avctx, codec_caps, &enccfg, &flags, &img_fmt))	907	if (set_pix_fmt(avctx, codec_caps, &enccfg, &flags, &img_fmt))
853	return AVERROR(EINVAL);	908	return AVERROR(EINVAL);
		909	// Keep HDR10+ if it has bit depth higher than 8 and
		910	// it has PQ trc (SMPTE2084).
		911	if (enccfg.g_bit_depth > 8 && avctx->color_trc == AVCOL_TRC_SMPTE2084) {
		912	ctx->discard_hdr10_plus = 0;
		913	ctx->hdr10_plus_fifo = av_fifo_alloc(sizeof(FrameHDR10Plus));
		914	if (!ctx->hdr10_plus_fifo)
		915	return AVERROR(ENOMEM);
		916	}
854	}	917	}
855	#endif	918	#endif
856		919
@@ -1213,6 +1276,15 @@ static int storeframe(AVCodecContext avctx, struct FrameListData cx_frame, · 356 unmodified lines
1213	AV_WB64(side_data, 1);	1276	AV_WB64(side_data, 1);
1214	memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);	1277	memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
1215	}	1278	}
		1279	if (cx_frame->frame_number != -1) {
		1280	VPxContext *ctx = avctx->priv_data;
		1281	if (!ctx->discard_hdr10_plus) {
		1282	int err = copy_hdr10_plus_to_pkt(ctx->hdr10_plus_fifo, pkt);
		1283	if (err < 0)
		1284	return err;
		1285	}
		1286	}
		1287
1216	return pkt->size;	1288	return pkt->size;
1217	}	1289	}
1218		1290
@@ -1620,6 +1692,25 @@ static int vpx_encode(AVCodecContext avctx, AVPacket pkt, · 401 unmodified lines
1620	vp9_encode_set_roi(avctx, frame->width, frame->height, sd);	1692	vp9_encode_set_roi(avctx, frame->width, frame->height, sd);
1621	}	1693	}
1622	}	1694	}
		1695
		1696	if (!ctx->discard_hdr10_plus) {
		1697	AVFrameSideData *hdr10_plus_metadata;
		1698	// Add HDR10+ metadata to queue.
		1699	hdr10_plus_metadata = av_frame_get_side_data(frame, AV_FRAME_DATA_DYNAMIC_HDR_PLUS);
		1700	if (hdr10_plus_metadata) {
		1701	int err;
		1702	struct FrameHDR10Plus data;
		1703	data.pts = frame->pts;
		1704	data.hdr10_plus = av_buffer_ref(hdr10_plus_metadata->buf);
		1705	if (!data.hdr10_plus)
		1706	return AVERROR(ENOMEM);
		1707	err = add_hdr10_plus(ctx->hdr10_plus_fifo, &data);
		1708	if (err < 0) {
		1709	av_buffer_unref(&data.hdr10_plus);
		1710	return err;
		1711	}
		1712	}
		1713	}
1623	}	1714	}
1624		1715
1625	// this is for encoding with preset temporal layering patterns defined in	1716	// this is for encoding with preset temporal layering patterns defined in

libavcodec/packet.h+8−0

@@ -290,6 +290,14 @@ enum AVPacketSideDataType { · 289 unmodified lines
290	*/	290	*/
291	AV_PKT_DATA_S12M_TIMECODE,	291	AV_PKT_DATA_S12M_TIMECODE,
292		292
		293	/**
		294	* HDR10+ dynamic metadata associated with a video frame. The metadata is in
		295	* the form of the AVDynamicHDRPlus struct and contains
		296	* information for color volume transform - application 4 of
		297	* SMPTE 2094-40:2016 standard.
		298	*/
		299	AV_PKT_DATA_DYNAMIC_HDR10_PLUS,
		300
293	/**	301	/**
294	* The number of side data types.	302	* The number of side data types.
295	* This is not part of the public API/ABI in the sense that it may	303	* This is not part of the public API/ABI in the sense that it may

libavcodec/version.h+2−2

@@ -28,8 +28,8 @@ · 27 unmodified lines
28	#include "libavutil/version.h"	28	#include "libavutil/version.h"
29		29
30	#define LIBAVCODEC_VERSION_MAJOR 59	30	#define LIBAVCODEC_VERSION_MAJOR 59
31	#define LIBAVCODEC_VERSION_MINOR 1	31	#define LIBAVCODEC_VERSION_MINOR 2
32	#define LIBAVCODEC_VERSION_MICRO 102	32	#define LIBAVCODEC_VERSION_MICRO 100
33		33
34	#define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \	34	#define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
35	LIBAVCODEC_VERSION_MINOR, \	35	LIBAVCODEC_VERSION_MINOR, \

avcodec: Pass HDR10+ metadata to packet side data in VP9 encoder

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code