2 * This file is part of FFmpeg.
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 #include "libavutil/common.h"
20 #include "libavutil/opt.h"
22 #include "bsf_internal.h"
26 #include "h265_profile_level.h"
39 typedef struct H265MetadataContext {
42 CodedBitstreamContext *cbc;
43 CodedBitstreamFragment access_unit;
49 AVRational sample_aspect_ratio;
52 int video_full_range_flag;
54 int transfer_characteristics;
55 int matrix_coefficients;
57 int chroma_sample_loc_type;
60 int poc_proportional_to_timing_flag;
61 int num_ticks_poc_diff_one;
71 } H265MetadataContext;
74 static void h265_metadata_guess_level(AVBSFContext *bsf,
75 const CodedBitstreamFragment *au)
77 H265MetadataContext *ctx = bsf->priv_data;
78 const H265LevelDescriptor *desc;
79 const H265RawProfileTierLevel *ptl = NULL;
80 const H265RawHRDParameters *hrd = NULL;
82 int width = 0, height = 0;
83 int tile_cols = 0, tile_rows = 0;
84 int max_dec_pic_buffering = 0;
87 for (i = 0; i < au->nb_units; i++) {
88 const CodedBitstreamUnit *unit = &au->units[i];
90 if (unit->type == HEVC_NAL_VPS) {
91 const H265RawVPS *vps = unit->content;
93 ptl = &vps->profile_tier_level;
94 max_dec_pic_buffering = vps->vps_max_dec_pic_buffering_minus1[0] + 1;
96 if (vps->vps_num_hrd_parameters > 0)
97 hrd = &vps->hrd_parameters[0];
99 } else if (unit->type == HEVC_NAL_SPS) {
100 const H265RawSPS *sps = unit->content;
102 ptl = &sps->profile_tier_level;
103 max_dec_pic_buffering = sps->sps_max_dec_pic_buffering_minus1[0] + 1;
105 width = sps->pic_width_in_luma_samples;
106 height = sps->pic_height_in_luma_samples;
108 if (sps->vui.vui_hrd_parameters_present_flag)
109 hrd = &sps->vui.hrd_parameters;
111 } else if (unit->type == HEVC_NAL_PPS) {
112 const H265RawPPS *pps = unit->content;
114 if (pps->tiles_enabled_flag) {
115 tile_cols = pps->num_tile_columns_minus1 + 1;
116 tile_rows = pps->num_tile_rows_minus1 + 1;
122 if (hrd->nal_hrd_parameters_present_flag) {
123 bit_rate = (hrd->nal_sub_layer_hrd_parameters[0].bit_rate_value_minus1[0] + 1) *
124 (INT64_C(1) << hrd->bit_rate_scale + 6);
125 } else if (hrd->vcl_hrd_parameters_present_flag) {
126 bit_rate = (hrd->vcl_sub_layer_hrd_parameters[0].bit_rate_value_minus1[0] + 1) *
127 (INT64_C(1) << hrd->bit_rate_scale + 6);
128 // Adjust for VCL vs. NAL limits.
129 bit_rate = bit_rate * 11 / 10;
133 desc = ff_h265_guess_level(ptl, bit_rate, width, height,
134 0, tile_rows, tile_cols,
135 max_dec_pic_buffering);
137 av_log(bsf, AV_LOG_DEBUG, "Stream appears to conform to "
138 "level %s.\n", desc->name);
139 ctx->level_guess = desc->level_idc;
143 static void h265_metadata_update_level(AVBSFContext *bsf,
146 H265MetadataContext *ctx = bsf->priv_data;
148 if (ctx->level != LEVEL_UNSET) {
149 if (ctx->level == LEVEL_AUTO) {
150 if (ctx->level_guess) {
151 *level_idc = ctx->level_guess;
153 if (!ctx->level_warned) {
154 av_log(bsf, AV_LOG_WARNING, "Unable to determine level "
155 "of stream: using level 8.5.\n");
156 ctx->level_warned = 1;
161 *level_idc = ctx->level;
166 static int h265_metadata_update_vps(AVBSFContext *bsf,
169 H265MetadataContext *ctx = bsf->priv_data;
171 if (ctx->tick_rate.num && ctx->tick_rate.den) {
174 av_reduce(&num, &den, ctx->tick_rate.num, ctx->tick_rate.den,
175 UINT32_MAX > INT_MAX ? UINT32_MAX : INT_MAX);
177 vps->vps_time_scale = num;
178 vps->vps_num_units_in_tick = den;
180 vps->vps_timing_info_present_flag = 1;
182 if (ctx->num_ticks_poc_diff_one > 0) {
183 vps->vps_num_ticks_poc_diff_one_minus1 =
184 ctx->num_ticks_poc_diff_one - 1;
185 vps->vps_poc_proportional_to_timing_flag = 1;
186 } else if (ctx->num_ticks_poc_diff_one == 0) {
187 vps->vps_poc_proportional_to_timing_flag = 0;
191 h265_metadata_update_level(bsf, &vps->profile_tier_level.general_level_idc);
196 static int h265_metadata_update_sps(AVBSFContext *bsf,
199 H265MetadataContext *ctx = bsf->priv_data;
201 int crop_unit_x, crop_unit_y;
203 if (ctx->sample_aspect_ratio.num && ctx->sample_aspect_ratio.den) {
205 static const AVRational sar_idc[] = {
206 { 0, 0 }, // Unspecified (never written here).
207 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
208 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
209 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
210 { 160, 99 }, { 4, 3 }, { 3, 2 }, { 2, 1 },
214 av_reduce(&num, &den, ctx->sample_aspect_ratio.num,
215 ctx->sample_aspect_ratio.den, 65535);
217 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
218 if (num == sar_idc[i].num &&
219 den == sar_idc[i].den)
222 if (i == FF_ARRAY_ELEMS(sar_idc)) {
223 sps->vui.aspect_ratio_idc = 255;
224 sps->vui.sar_width = num;
225 sps->vui.sar_height = den;
227 sps->vui.aspect_ratio_idc = i;
229 sps->vui.aspect_ratio_info_present_flag = 1;
233 #define SET_OR_INFER(field, value, present_flag, infer) do { \
237 } else if (!present_flag) \
241 if (ctx->video_format >= 0 ||
242 ctx->video_full_range_flag >= 0 ||
243 ctx->colour_primaries >= 0 ||
244 ctx->transfer_characteristics >= 0 ||
245 ctx->matrix_coefficients >= 0) {
247 SET_OR_INFER(sps->vui.video_format, ctx->video_format,
248 sps->vui.video_signal_type_present_flag, 5);
250 SET_OR_INFER(sps->vui.video_full_range_flag,
251 ctx->video_full_range_flag,
252 sps->vui.video_signal_type_present_flag, 0);
254 if (ctx->colour_primaries >= 0 ||
255 ctx->transfer_characteristics >= 0 ||
256 ctx->matrix_coefficients >= 0) {
258 SET_OR_INFER(sps->vui.colour_primaries,
259 ctx->colour_primaries,
260 sps->vui.colour_description_present_flag, 2);
262 SET_OR_INFER(sps->vui.transfer_characteristics,
263 ctx->transfer_characteristics,
264 sps->vui.colour_description_present_flag, 2);
266 SET_OR_INFER(sps->vui.matrix_coefficients,
267 ctx->matrix_coefficients,
268 sps->vui.colour_description_present_flag, 2);
270 sps->vui.colour_description_present_flag = 1;
272 sps->vui.video_signal_type_present_flag = 1;
276 if (ctx->chroma_sample_loc_type >= 0) {
277 sps->vui.chroma_sample_loc_type_top_field =
278 ctx->chroma_sample_loc_type;
279 sps->vui.chroma_sample_loc_type_bottom_field =
280 ctx->chroma_sample_loc_type;
281 sps->vui.chroma_loc_info_present_flag = 1;
285 if (ctx->tick_rate.num && ctx->tick_rate.den) {
288 av_reduce(&num, &den, ctx->tick_rate.num, ctx->tick_rate.den,
289 UINT32_MAX > INT_MAX ? UINT32_MAX : INT_MAX);
291 sps->vui.vui_time_scale = num;
292 sps->vui.vui_num_units_in_tick = den;
294 sps->vui.vui_timing_info_present_flag = 1;
297 if (ctx->num_ticks_poc_diff_one > 0) {
298 sps->vui.vui_num_ticks_poc_diff_one_minus1 =
299 ctx->num_ticks_poc_diff_one - 1;
300 sps->vui.vui_poc_proportional_to_timing_flag = 1;
301 } else if (ctx->num_ticks_poc_diff_one == 0) {
302 sps->vui.vui_poc_proportional_to_timing_flag = 0;
306 if (sps->separate_colour_plane_flag || sps->chroma_format_idc == 0) {
310 crop_unit_x = 1 + (sps->chroma_format_idc < 3);
311 crop_unit_y = 1 + (sps->chroma_format_idc < 2);
313 #define CROP(border, unit) do { \
314 if (ctx->crop_ ## border >= 0) { \
315 if (ctx->crop_ ## border % unit != 0) { \
316 av_log(bsf, AV_LOG_ERROR, "Invalid value for crop_%s: " \
317 "must be a multiple of %d.\n", #border, unit); \
318 return AVERROR(EINVAL); \
320 sps->conf_win_ ## border ## _offset = \
321 ctx->crop_ ## border / unit; \
322 sps->conformance_window_flag = 1; \
325 CROP(left, crop_unit_x);
326 CROP(right, crop_unit_x);
327 CROP(top, crop_unit_y);
328 CROP(bottom, crop_unit_y);
332 sps->vui_parameters_present_flag = 1;
334 h265_metadata_update_level(bsf, &sps->profile_tier_level.general_level_idc);
339 static int h265_metadata_update_side_data(AVBSFContext *bsf, AVPacket *pkt)
341 H265MetadataContext *ctx = bsf->priv_data;
342 CodedBitstreamFragment *au = &ctx->access_unit;
347 side_data = av_packet_get_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA,
352 err = ff_cbs_read(ctx->cbc, au, side_data, side_data_size);
354 av_log(bsf, AV_LOG_ERROR, "Failed to read extradata from packet side data.\n");
358 if (ctx->level == LEVEL_AUTO && !ctx->level_guess)
359 h265_metadata_guess_level(bsf, au);
361 for (i = 0; i < au->nb_units; i++) {
362 if (au->units[i].type == HEVC_NAL_VPS) {
363 err = h265_metadata_update_vps(bsf, au->units[i].content);
367 if (au->units[i].type == HEVC_NAL_SPS) {
368 err = h265_metadata_update_sps(bsf, au->units[i].content);
374 err = ff_cbs_write_fragment_data(ctx->cbc, au);
376 av_log(bsf, AV_LOG_ERROR, "Failed to write extradata into packet side data.\n");
380 side_data = av_packet_new_side_data(pkt, AV_PKT_DATA_NEW_EXTRADATA, au->data_size);
382 return AVERROR(ENOMEM);
383 memcpy(side_data, au->data, au->data_size);
385 ff_cbs_fragment_reset(ctx->cbc, au);
390 static int h265_metadata_filter(AVBSFContext *bsf, AVPacket *pkt)
392 H265MetadataContext *ctx = bsf->priv_data;
393 CodedBitstreamFragment *au = &ctx->access_unit;
396 err = ff_bsf_get_packet_ref(bsf, pkt);
400 err = h265_metadata_update_side_data(bsf, pkt);
404 err = ff_cbs_read_packet(ctx->cbc, au, pkt);
406 av_log(bsf, AV_LOG_ERROR, "Failed to read packet.\n");
410 if (au->nb_units == 0) {
411 av_log(bsf, AV_LOG_ERROR, "No NAL units in packet.\n");
412 err = AVERROR_INVALIDDATA;
416 // If an AUD is present, it must be the first NAL unit.
417 if (au->units[0].type == HEVC_NAL_AUD) {
418 if (ctx->aud == REMOVE)
419 ff_cbs_delete_unit(ctx->cbc, au, 0);
421 if (ctx->aud == INSERT) {
422 H265RawAUD *aud = &ctx->aud_nal;
423 int pic_type = 0, temporal_id = 8, layer_id = 0;
425 for (i = 0; i < au->nb_units; i++) {
426 const H265RawNALUnitHeader *nal = au->units[i].content;
429 if (nal->nuh_temporal_id_plus1 < temporal_id + 1)
430 temporal_id = nal->nuh_temporal_id_plus1 - 1;
432 if (au->units[i].type <= HEVC_NAL_RSV_VCL31) {
433 const H265RawSlice *slice = au->units[i].content;
434 layer_id = nal->nuh_layer_id;
435 if (slice->header.slice_type == HEVC_SLICE_B &&
438 if (slice->header.slice_type == HEVC_SLICE_P &&
444 aud->nal_unit_header = (H265RawNALUnitHeader) {
445 .nal_unit_type = HEVC_NAL_AUD,
446 .nuh_layer_id = layer_id,
447 .nuh_temporal_id_plus1 = temporal_id + 1,
449 aud->pic_type = pic_type;
451 err = ff_cbs_insert_unit_content(ctx->cbc, au,
452 0, HEVC_NAL_AUD, aud, NULL);
454 av_log(bsf, AV_LOG_ERROR, "Failed to insert AUD.\n");
460 if (ctx->level == LEVEL_AUTO && !ctx->level_guess)
461 h265_metadata_guess_level(bsf, au);
463 for (i = 0; i < au->nb_units; i++) {
464 if (au->units[i].type == HEVC_NAL_VPS) {
465 err = h265_metadata_update_vps(bsf, au->units[i].content);
469 if (au->units[i].type == HEVC_NAL_SPS) {
470 err = h265_metadata_update_sps(bsf, au->units[i].content);
476 err = ff_cbs_write_packet(ctx->cbc, pkt, au);
478 av_log(bsf, AV_LOG_ERROR, "Failed to write packet.\n");
484 ff_cbs_fragment_reset(ctx->cbc, au);
487 av_packet_unref(pkt);
492 static int h265_metadata_init(AVBSFContext *bsf)
494 H265MetadataContext *ctx = bsf->priv_data;
495 CodedBitstreamFragment *au = &ctx->access_unit;
498 err = ff_cbs_init(&ctx->cbc, AV_CODEC_ID_HEVC, bsf);
502 if (bsf->par_in->extradata) {
503 err = ff_cbs_read_extradata(ctx->cbc, au, bsf->par_in);
505 av_log(bsf, AV_LOG_ERROR, "Failed to read extradata.\n");
509 if (ctx->level == LEVEL_AUTO)
510 h265_metadata_guess_level(bsf, au);
512 for (i = 0; i < au->nb_units; i++) {
513 if (au->units[i].type == HEVC_NAL_VPS) {
514 err = h265_metadata_update_vps(bsf, au->units[i].content);
518 if (au->units[i].type == HEVC_NAL_SPS) {
519 err = h265_metadata_update_sps(bsf, au->units[i].content);
525 err = ff_cbs_write_extradata(ctx->cbc, bsf->par_out, au);
527 av_log(bsf, AV_LOG_ERROR, "Failed to write extradata.\n");
534 ff_cbs_fragment_reset(ctx->cbc, au);
538 static void h265_metadata_close(AVBSFContext *bsf)
540 H265MetadataContext *ctx = bsf->priv_data;
542 ff_cbs_fragment_free(ctx->cbc, &ctx->access_unit);
543 ff_cbs_close(&ctx->cbc);
546 #define OFFSET(x) offsetof(H265MetadataContext, x)
547 #define FLAGS (AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_BSF_PARAM)
548 static const AVOption h265_metadata_options[] = {
549 { "aud", "Access Unit Delimiter NAL units",
550 OFFSET(aud), AV_OPT_TYPE_INT,
551 { .i64 = PASS }, PASS, REMOVE, FLAGS, "aud" },
552 { "pass", NULL, 0, AV_OPT_TYPE_CONST,
553 { .i64 = PASS }, .flags = FLAGS, .unit = "aud" },
554 { "insert", NULL, 0, AV_OPT_TYPE_CONST,
555 { .i64 = INSERT }, .flags = FLAGS, .unit = "aud" },
556 { "remove", NULL, 0, AV_OPT_TYPE_CONST,
557 { .i64 = REMOVE }, .flags = FLAGS, .unit = "aud" },
559 { "sample_aspect_ratio", "Set sample aspect ratio (table E-1)",
560 OFFSET(sample_aspect_ratio), AV_OPT_TYPE_RATIONAL,
561 { .dbl = 0.0 }, 0, 65535, FLAGS },
563 { "video_format", "Set video format (table E-2)",
564 OFFSET(video_format), AV_OPT_TYPE_INT,
565 { .i64 = -1 }, -1, 7, FLAGS },
566 { "video_full_range_flag", "Set video full range flag",
567 OFFSET(video_full_range_flag), AV_OPT_TYPE_INT,
568 { .i64 = -1 }, -1, 1, FLAGS },
569 { "colour_primaries", "Set colour primaries (table E-3)",
570 OFFSET(colour_primaries), AV_OPT_TYPE_INT,
571 { .i64 = -1 }, -1, 255, FLAGS },
572 { "transfer_characteristics", "Set transfer characteristics (table E-4)",
573 OFFSET(transfer_characteristics), AV_OPT_TYPE_INT,
574 { .i64 = -1 }, -1, 255, FLAGS },
575 { "matrix_coefficients", "Set matrix coefficients (table E-5)",
576 OFFSET(matrix_coefficients), AV_OPT_TYPE_INT,
577 { .i64 = -1 }, -1, 255, FLAGS },
579 { "chroma_sample_loc_type", "Set chroma sample location type (figure E-1)",
580 OFFSET(chroma_sample_loc_type), AV_OPT_TYPE_INT,
581 { .i64 = -1 }, -1, 6, FLAGS },
584 "Set VPS and VUI tick rate (num_units_in_tick / time_scale)",
585 OFFSET(tick_rate), AV_OPT_TYPE_RATIONAL,
586 { .dbl = 0.0 }, 0, UINT_MAX, FLAGS },
587 { "num_ticks_poc_diff_one",
588 "Set VPS and VUI number of ticks per POC increment",
589 OFFSET(num_ticks_poc_diff_one), AV_OPT_TYPE_INT,
590 { .i64 = -1 }, -1, INT_MAX, FLAGS },
592 { "crop_left", "Set left border crop offset",
593 OFFSET(crop_left), AV_OPT_TYPE_INT,
594 { .i64 = -1 }, -1, HEVC_MAX_WIDTH, FLAGS },
595 { "crop_right", "Set right border crop offset",
596 OFFSET(crop_right), AV_OPT_TYPE_INT,
597 { .i64 = -1 }, -1, HEVC_MAX_WIDTH, FLAGS },
598 { "crop_top", "Set top border crop offset",
599 OFFSET(crop_top), AV_OPT_TYPE_INT,
600 { .i64 = -1 }, -1, HEVC_MAX_HEIGHT, FLAGS },
601 { "crop_bottom", "Set bottom border crop offset",
602 OFFSET(crop_bottom), AV_OPT_TYPE_INT,
603 { .i64 = -1 }, -1, HEVC_MAX_HEIGHT, FLAGS },
605 { "level", "Set level (tables A.6 and A.7)",
606 OFFSET(level), AV_OPT_TYPE_INT,
607 { .i64 = LEVEL_UNSET }, LEVEL_UNSET, 0xff, FLAGS, "level" },
608 { "auto", "Attempt to guess level from stream properties",
609 0, AV_OPT_TYPE_CONST,
610 { .i64 = LEVEL_AUTO }, .flags = FLAGS, .unit = "level" },
611 #define LEVEL(name, value) name, NULL, 0, AV_OPT_TYPE_CONST, \
612 { .i64 = value }, .flags = FLAGS, .unit = "level"
615 { LEVEL("2.1", 63) },
617 { LEVEL("3.1", 93) },
619 { LEVEL("4.1", 123) },
621 { LEVEL("5.1", 153) },
622 { LEVEL("5.2", 156) },
624 { LEVEL("6.1", 183) },
625 { LEVEL("6.2", 186) },
626 { LEVEL("8.5", 255) },
632 static const AVClass h265_metadata_class = {
633 .class_name = "h265_metadata_bsf",
634 .item_name = av_default_item_name,
635 .option = h265_metadata_options,
636 .version = LIBAVUTIL_VERSION_INT,
639 static const enum AVCodecID h265_metadata_codec_ids[] = {
640 AV_CODEC_ID_HEVC, AV_CODEC_ID_NONE,
643 const AVBitStreamFilter ff_hevc_metadata_bsf = {
644 .name = "hevc_metadata",
645 .priv_data_size = sizeof(H265MetadataContext),
646 .priv_class = &h265_metadata_class,
647 .init = &h265_metadata_init,
648 .close = &h265_metadata_close,
649 .filter = &h265_metadata_filter,
650 .codec_ids = h265_metadata_codec_ids,