1 /*
2 * This file is part of FFmpeg.
3 *
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.
8 *
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.
13 *
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
17 */
18
19 #include "libavutil/common.h"
20 #include "libavutil/opt.h"
21
22 #include "bsf.h"
23 #include "bsf_internal.h"
24 #include "cbs.h"
25 #include "cbs_bsf.h"
26 #include "cbs_h265.h"
27 #include "hevc.h"
28 #include "h265_profile_level.h"
29
30 enum {
31 LEVEL_UNSET = -2,
32 LEVEL_AUTO = -1,
33 };
34
35 typedef struct H265MetadataContext {
36 CBSBSFContext common;
37
38 H265RawAUD aud_nal;
39
40 int aud;
41
42 AVRational sample_aspect_ratio;
43
44 int video_format;
45 int video_full_range_flag;
46 int colour_primaries;
47 int transfer_characteristics;
48 int matrix_coefficients;
49
50 int chroma_sample_loc_type;
51
52 AVRational tick_rate;
53 int poc_proportional_to_timing_flag;
54 int num_ticks_poc_diff_one;
55
56 int crop_left;
57 int crop_right;
58 int crop_top;
59 int crop_bottom;
60
61 int level;
62 int level_guess;
63 int level_warned;
64 } H265MetadataContext;
65
66
h265_metadata_guess_level(AVBSFContext *bsf, const CodedBitstreamFragment *au)67 static void h265_metadata_guess_level(AVBSFContext *bsf,
68 const CodedBitstreamFragment *au)
69 {
70 H265MetadataContext *ctx = bsf->priv_data;
71 const H265LevelDescriptor *desc;
72 const H265RawProfileTierLevel *ptl = NULL;
73 const H265RawHRDParameters *hrd = NULL;
74 int64_t bit_rate = 0;
75 int width = 0, height = 0;
76 int tile_cols = 0, tile_rows = 0;
77 int max_dec_pic_buffering = 0;
78 int i;
79
80 for (i = 0; i < au->nb_units; i++) {
81 const CodedBitstreamUnit *unit = &au->units[i];
82
83 if (unit->type == HEVC_NAL_VPS) {
84 const H265RawVPS *vps = unit->content;
85
86 ptl = &vps->profile_tier_level;
87 max_dec_pic_buffering = vps->vps_max_dec_pic_buffering_minus1[0] + 1;
88
89 if (vps->vps_num_hrd_parameters > 0)
90 hrd = &vps->hrd_parameters[0];
91
92 } else if (unit->type == HEVC_NAL_SPS) {
93 const H265RawSPS *sps = unit->content;
94
95 ptl = &sps->profile_tier_level;
96 max_dec_pic_buffering = sps->sps_max_dec_pic_buffering_minus1[0] + 1;
97
98 width = sps->pic_width_in_luma_samples;
99 height = sps->pic_height_in_luma_samples;
100
101 if (sps->vui.vui_hrd_parameters_present_flag)
102 hrd = &sps->vui.hrd_parameters;
103
104 } else if (unit->type == HEVC_NAL_PPS) {
105 const H265RawPPS *pps = unit->content;
106
107 if (pps->tiles_enabled_flag) {
108 tile_cols = pps->num_tile_columns_minus1 + 1;
109 tile_rows = pps->num_tile_rows_minus1 + 1;
110 }
111 }
112 }
113
114 if (hrd) {
115 if (hrd->nal_hrd_parameters_present_flag) {
116 bit_rate = (hrd->nal_sub_layer_hrd_parameters[0].bit_rate_value_minus1[0] + 1) *
117 (INT64_C(1) << hrd->bit_rate_scale + 6);
118 } else if (hrd->vcl_hrd_parameters_present_flag) {
119 bit_rate = (hrd->vcl_sub_layer_hrd_parameters[0].bit_rate_value_minus1[0] + 1) *
120 (INT64_C(1) << hrd->bit_rate_scale + 6);
121 // Adjust for VCL vs. NAL limits.
122 bit_rate = bit_rate * 11 / 10;
123 }
124 }
125
126 desc = ff_h265_guess_level(ptl, bit_rate, width, height,
127 0, tile_rows, tile_cols,
128 max_dec_pic_buffering);
129 if (desc) {
130 av_log(bsf, AV_LOG_DEBUG, "Stream appears to conform to "
131 "level %s.\n", desc->name);
132 ctx->level_guess = desc->level_idc;
133 }
134 }
135
h265_metadata_update_level(AVBSFContext *bsf, uint8_t *level_idc)136 static void h265_metadata_update_level(AVBSFContext *bsf,
137 uint8_t *level_idc)
138 {
139 H265MetadataContext *ctx = bsf->priv_data;
140
141 if (ctx->level != LEVEL_UNSET) {
142 if (ctx->level == LEVEL_AUTO) {
143 if (ctx->level_guess) {
144 *level_idc = ctx->level_guess;
145 } else {
146 if (!ctx->level_warned) {
147 av_log(bsf, AV_LOG_WARNING, "Unable to determine level "
148 "of stream: using level 8.5.\n");
149 ctx->level_warned = 1;
150 }
151 *level_idc = 255;
152 }
153 } else {
154 *level_idc = ctx->level;
155 }
156 }
157 }
158
h265_metadata_update_vps(AVBSFContext *bsf, H265RawVPS *vps)159 static int h265_metadata_update_vps(AVBSFContext *bsf,
160 H265RawVPS *vps)
161 {
162 H265MetadataContext *ctx = bsf->priv_data;
163
164 if (ctx->tick_rate.num && ctx->tick_rate.den) {
165 int num, den;
166
167 av_reduce(&num, &den, ctx->tick_rate.num, ctx->tick_rate.den,
168 UINT32_MAX > INT_MAX ? UINT32_MAX : INT_MAX);
169
170 vps->vps_time_scale = num;
171 vps->vps_num_units_in_tick = den;
172
173 vps->vps_timing_info_present_flag = 1;
174
175 if (ctx->num_ticks_poc_diff_one > 0) {
176 vps->vps_num_ticks_poc_diff_one_minus1 =
177 ctx->num_ticks_poc_diff_one - 1;
178 vps->vps_poc_proportional_to_timing_flag = 1;
179 } else if (ctx->num_ticks_poc_diff_one == 0) {
180 vps->vps_poc_proportional_to_timing_flag = 0;
181 }
182 }
183
184 h265_metadata_update_level(bsf, &vps->profile_tier_level.general_level_idc);
185
186 return 0;
187 }
188
h265_metadata_update_sps(AVBSFContext *bsf, H265RawSPS *sps)189 static int h265_metadata_update_sps(AVBSFContext *bsf,
190 H265RawSPS *sps)
191 {
192 H265MetadataContext *ctx = bsf->priv_data;
193 int need_vui = 0;
194 int crop_unit_x, crop_unit_y;
195
196 if (ctx->sample_aspect_ratio.num && ctx->sample_aspect_ratio.den) {
197 // Table E-1.
198 static const AVRational sar_idc[] = {
199 { 0, 0 }, // Unspecified (never written here).
200 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
201 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
202 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
203 { 160, 99 }, { 4, 3 }, { 3, 2 }, { 2, 1 },
204 };
205 int num, den, i;
206
207 av_reduce(&num, &den, ctx->sample_aspect_ratio.num,
208 ctx->sample_aspect_ratio.den, 65535);
209
210 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
211 if (num == sar_idc[i].num &&
212 den == sar_idc[i].den)
213 break;
214 }
215 if (i == FF_ARRAY_ELEMS(sar_idc)) {
216 sps->vui.aspect_ratio_idc = 255;
217 sps->vui.sar_width = num;
218 sps->vui.sar_height = den;
219 } else {
220 sps->vui.aspect_ratio_idc = i;
221 }
222 sps->vui.aspect_ratio_info_present_flag = 1;
223 need_vui = 1;
224 }
225
226 #define SET_OR_INFER(field, value, present_flag, infer) do { \
227 if (value >= 0) { \
228 field = value; \
229 need_vui = 1; \
230 } else if (!present_flag) \
231 field = infer; \
232 } while (0)
233
234 if (ctx->video_format >= 0 ||
235 ctx->video_full_range_flag >= 0 ||
236 ctx->colour_primaries >= 0 ||
237 ctx->transfer_characteristics >= 0 ||
238 ctx->matrix_coefficients >= 0) {
239
240 SET_OR_INFER(sps->vui.video_format, ctx->video_format,
241 sps->vui.video_signal_type_present_flag, 5);
242
243 SET_OR_INFER(sps->vui.video_full_range_flag,
244 ctx->video_full_range_flag,
245 sps->vui.video_signal_type_present_flag, 0);
246
247 if (ctx->colour_primaries >= 0 ||
248 ctx->transfer_characteristics >= 0 ||
249 ctx->matrix_coefficients >= 0) {
250
251 SET_OR_INFER(sps->vui.colour_primaries,
252 ctx->colour_primaries,
253 sps->vui.colour_description_present_flag, 2);
254
255 SET_OR_INFER(sps->vui.transfer_characteristics,
256 ctx->transfer_characteristics,
257 sps->vui.colour_description_present_flag, 2);
258
259 SET_OR_INFER(sps->vui.matrix_coefficients,
260 ctx->matrix_coefficients,
261 sps->vui.colour_description_present_flag, 2);
262
263 sps->vui.colour_description_present_flag = 1;
264 }
265 sps->vui.video_signal_type_present_flag = 1;
266 need_vui = 1;
267 }
268
269 if (ctx->chroma_sample_loc_type >= 0) {
270 sps->vui.chroma_sample_loc_type_top_field =
271 ctx->chroma_sample_loc_type;
272 sps->vui.chroma_sample_loc_type_bottom_field =
273 ctx->chroma_sample_loc_type;
274 sps->vui.chroma_loc_info_present_flag = 1;
275 need_vui = 1;
276 }
277
278 if (ctx->tick_rate.num && ctx->tick_rate.den) {
279 int num, den;
280
281 av_reduce(&num, &den, ctx->tick_rate.num, ctx->tick_rate.den,
282 UINT32_MAX > INT_MAX ? UINT32_MAX : INT_MAX);
283
284 sps->vui.vui_time_scale = num;
285 sps->vui.vui_num_units_in_tick = den;
286
287 sps->vui.vui_timing_info_present_flag = 1;
288 need_vui = 1;
289
290 if (ctx->num_ticks_poc_diff_one > 0) {
291 sps->vui.vui_num_ticks_poc_diff_one_minus1 =
292 ctx->num_ticks_poc_diff_one - 1;
293 sps->vui.vui_poc_proportional_to_timing_flag = 1;
294 } else if (ctx->num_ticks_poc_diff_one == 0) {
295 sps->vui.vui_poc_proportional_to_timing_flag = 0;
296 }
297 }
298
299 if (sps->separate_colour_plane_flag || sps->chroma_format_idc == 0) {
300 crop_unit_x = 1;
301 crop_unit_y = 1;
302 } else {
303 crop_unit_x = 1 + (sps->chroma_format_idc < 3);
304 crop_unit_y = 1 + (sps->chroma_format_idc < 2);
305 }
306 #define CROP(border, unit) do { \
307 if (ctx->crop_ ## border >= 0) { \
308 if (ctx->crop_ ## border % unit != 0) { \
309 av_log(bsf, AV_LOG_ERROR, "Invalid value for crop_%s: " \
310 "must be a multiple of %d.\n", #border, unit); \
311 return AVERROR(EINVAL); \
312 } \
313 sps->conf_win_ ## border ## _offset = \
314 ctx->crop_ ## border / unit; \
315 sps->conformance_window_flag = 1; \
316 } \
317 } while (0)
318 CROP(left, crop_unit_x);
319 CROP(right, crop_unit_x);
320 CROP(top, crop_unit_y);
321 CROP(bottom, crop_unit_y);
322 #undef CROP
323
324 if (need_vui)
325 sps->vui_parameters_present_flag = 1;
326
327 h265_metadata_update_level(bsf, &sps->profile_tier_level.general_level_idc);
328
329 return 0;
330 }
331
h265_metadata_update_fragment(AVBSFContext *bsf, AVPacket *pkt, CodedBitstreamFragment *au)332 static int h265_metadata_update_fragment(AVBSFContext *bsf, AVPacket *pkt,
333 CodedBitstreamFragment *au)
334 {
335 H265MetadataContext *ctx = bsf->priv_data;
336 int err, i;
337
338 // If an AUD is present, it must be the first NAL unit.
339 if (au->nb_units && au->units[0].type == HEVC_NAL_AUD) {
340 if (ctx->aud == BSF_ELEMENT_REMOVE)
341 ff_cbs_delete_unit(au, 0);
342 } else {
343 if (pkt && ctx->aud == BSF_ELEMENT_INSERT) {
344 H265RawAUD *aud = &ctx->aud_nal;
345 int pic_type = 0, temporal_id = 8, layer_id = 0;
346
347 for (i = 0; i < au->nb_units; i++) {
348 const H265RawNALUnitHeader *nal = au->units[i].content;
349 if (!nal)
350 continue;
351 if (nal->nuh_temporal_id_plus1 < temporal_id + 1)
352 temporal_id = nal->nuh_temporal_id_plus1 - 1;
353
354 if (au->units[i].type <= HEVC_NAL_RSV_VCL31) {
355 const H265RawSlice *slice = au->units[i].content;
356 layer_id = nal->nuh_layer_id;
357 if (slice->header.slice_type == HEVC_SLICE_B &&
358 pic_type < 2)
359 pic_type = 2;
360 if (slice->header.slice_type == HEVC_SLICE_P &&
361 pic_type < 1)
362 pic_type = 1;
363 }
364 }
365
366 aud->nal_unit_header = (H265RawNALUnitHeader) {
367 .nal_unit_type = HEVC_NAL_AUD,
368 .nuh_layer_id = layer_id,
369 .nuh_temporal_id_plus1 = temporal_id + 1,
370 };
371 aud->pic_type = pic_type;
372
373 err = ff_cbs_insert_unit_content(au, 0, HEVC_NAL_AUD, aud, NULL);
374 if (err < 0) {
375 av_log(bsf, AV_LOG_ERROR, "Failed to insert AUD.\n");
376 return err;
377 }
378 }
379 }
380
381 if (ctx->level == LEVEL_AUTO && !ctx->level_guess)
382 h265_metadata_guess_level(bsf, au);
383
384 for (i = 0; i < au->nb_units; i++) {
385 if (au->units[i].type == HEVC_NAL_VPS) {
386 err = h265_metadata_update_vps(bsf, au->units[i].content);
387 if (err < 0)
388 return err;
389 }
390 if (au->units[i].type == HEVC_NAL_SPS) {
391 err = h265_metadata_update_sps(bsf, au->units[i].content);
392 if (err < 0)
393 return err;
394 }
395 }
396
397 return 0;
398 }
399
400 static const CBSBSFType h265_metadata_type = {
401 .codec_id = AV_CODEC_ID_HEVC,
402 .fragment_name = "access unit",
403 .unit_name = "NAL unit",
404 .update_fragment = &h265_metadata_update_fragment,
405 };
406
h265_metadata_init(AVBSFContext *bsf)407 static int h265_metadata_init(AVBSFContext *bsf)
408 {
409 return ff_cbs_bsf_generic_init(bsf, &h265_metadata_type);
410 }
411
412 #define OFFSET(x) offsetof(H265MetadataContext, x)
413 #define FLAGS (AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_BSF_PARAM)
414 static const AVOption h265_metadata_options[] = {
415 BSF_ELEMENT_OPTIONS_PIR("aud", "Access Unit Delimiter NAL units",
416 aud, FLAGS),
417
418 { "sample_aspect_ratio", "Set sample aspect ratio (table E-1)",
419 OFFSET(sample_aspect_ratio), AV_OPT_TYPE_RATIONAL,
420 { .dbl = 0.0 }, 0, 65535, FLAGS },
421
422 { "video_format", "Set video format (table E-2)",
423 OFFSET(video_format), AV_OPT_TYPE_INT,
424 { .i64 = -1 }, -1, 7, FLAGS },
425 { "video_full_range_flag", "Set video full range flag",
426 OFFSET(video_full_range_flag), AV_OPT_TYPE_INT,
427 { .i64 = -1 }, -1, 1, FLAGS },
428 { "colour_primaries", "Set colour primaries (table E-3)",
429 OFFSET(colour_primaries), AV_OPT_TYPE_INT,
430 { .i64 = -1 }, -1, 255, FLAGS },
431 { "transfer_characteristics", "Set transfer characteristics (table E-4)",
432 OFFSET(transfer_characteristics), AV_OPT_TYPE_INT,
433 { .i64 = -1 }, -1, 255, FLAGS },
434 { "matrix_coefficients", "Set matrix coefficients (table E-5)",
435 OFFSET(matrix_coefficients), AV_OPT_TYPE_INT,
436 { .i64 = -1 }, -1, 255, FLAGS },
437
438 { "chroma_sample_loc_type", "Set chroma sample location type (figure E-1)",
439 OFFSET(chroma_sample_loc_type), AV_OPT_TYPE_INT,
440 { .i64 = -1 }, -1, 5, FLAGS },
441
442 { "tick_rate",
443 "Set VPS and VUI tick rate (time_scale / num_units_in_tick)",
444 OFFSET(tick_rate), AV_OPT_TYPE_RATIONAL,
445 { .dbl = 0.0 }, 0, UINT_MAX, FLAGS },
446 { "num_ticks_poc_diff_one",
447 "Set VPS and VUI number of ticks per POC increment",
448 OFFSET(num_ticks_poc_diff_one), AV_OPT_TYPE_INT,
449 { .i64 = -1 }, -1, INT_MAX, FLAGS },
450
451 { "crop_left", "Set left border crop offset",
452 OFFSET(crop_left), AV_OPT_TYPE_INT,
453 { .i64 = -1 }, -1, HEVC_MAX_WIDTH, FLAGS },
454 { "crop_right", "Set right border crop offset",
455 OFFSET(crop_right), AV_OPT_TYPE_INT,
456 { .i64 = -1 }, -1, HEVC_MAX_WIDTH, FLAGS },
457 { "crop_top", "Set top border crop offset",
458 OFFSET(crop_top), AV_OPT_TYPE_INT,
459 { .i64 = -1 }, -1, HEVC_MAX_HEIGHT, FLAGS },
460 { "crop_bottom", "Set bottom border crop offset",
461 OFFSET(crop_bottom), AV_OPT_TYPE_INT,
462 { .i64 = -1 }, -1, HEVC_MAX_HEIGHT, FLAGS },
463
464 { "level", "Set level (tables A.6 and A.7)",
465 OFFSET(level), AV_OPT_TYPE_INT,
466 { .i64 = LEVEL_UNSET }, LEVEL_UNSET, 0xff, FLAGS, "level" },
467 { "auto", "Attempt to guess level from stream properties",
468 0, AV_OPT_TYPE_CONST,
469 { .i64 = LEVEL_AUTO }, .flags = FLAGS, .unit = "level" },
470 #define LEVEL(name, value) name, NULL, 0, AV_OPT_TYPE_CONST, \
471 { .i64 = value }, .flags = FLAGS, .unit = "level"
472 { LEVEL("1", 30) },
473 { LEVEL("2", 60) },
474 { LEVEL("2.1", 63) },
475 { LEVEL("3", 90) },
476 { LEVEL("3.1", 93) },
477 { LEVEL("4", 120) },
478 { LEVEL("4.1", 123) },
479 { LEVEL("5", 150) },
480 { LEVEL("5.1", 153) },
481 { LEVEL("5.2", 156) },
482 { LEVEL("6", 180) },
483 { LEVEL("6.1", 183) },
484 { LEVEL("6.2", 186) },
485 { LEVEL("8.5", 255) },
486 #undef LEVEL
487
488 { NULL }
489 };
490
491 static const AVClass h265_metadata_class = {
492 .class_name = "h265_metadata_bsf",
493 .item_name = av_default_item_name,
494 .option = h265_metadata_options,
495 .version = LIBAVUTIL_VERSION_INT,
496 };
497
498 static const enum AVCodecID h265_metadata_codec_ids[] = {
499 AV_CODEC_ID_HEVC, AV_CODEC_ID_NONE,
500 };
501
502 const FFBitStreamFilter ff_hevc_metadata_bsf = {
503 .p.name = "hevc_metadata",
504 .p.codec_ids = h265_metadata_codec_ids,
505 .p.priv_class = &h265_metadata_class,
506 .priv_data_size = sizeof(H265MetadataContext),
507 .init = &h265_metadata_init,
508 .close = &ff_cbs_bsf_generic_close,
509 .filter = &ff_cbs_bsf_generic_filter,
510 };
511