lavc: Add hevc_metadata bitstream filter
[libav.git] / libavcodec / h265_metadata_bsf.c
1 /*
2 * This file is part of Libav.
3 *
4 * Libav 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 * Libav 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 Libav; 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 "cbs.h"
24 #include "cbs_h265.h"
25 #include "hevc.h"
26
27 enum {
28 PASS,
29 INSERT,
30 REMOVE,
31 };
32
33 typedef struct H265MetadataContext {
34 const AVClass *class;
35
36 CodedBitstreamContext cbc;
37 CodedBitstreamFragment access_unit;
38
39 H265RawAUD aud_nal;
40
41 int aud;
42
43 AVRational sample_aspect_ratio;
44
45 int video_format;
46 int video_full_range_flag;
47 int colour_primaries;
48 int transfer_characteristics;
49 int matrix_coefficients;
50
51 int chroma_sample_loc_type;
52
53 AVRational tick_rate;
54 int poc_proportional_to_timing_flag;
55 int num_ticks_poc_diff_one;
56
57 int crop_left;
58 int crop_right;
59 int crop_top;
60 int crop_bottom;
61 } H265MetadataContext;
62
63
64 static int h265_metadata_update_vps(AVBSFContext *bsf,
65 H265RawVPS *vps)
66 {
67 H265MetadataContext *ctx = bsf->priv_data;
68
69 if (ctx->tick_rate.num && ctx->tick_rate.den) {
70 int num, den;
71
72 av_reduce(&num, &den, ctx->tick_rate.num, ctx->tick_rate.den,
73 UINT32_MAX > INT_MAX ? UINT32_MAX : INT_MAX);
74
75 vps->vps_time_scale = num;
76 vps->vps_num_units_in_tick = den;
77
78 vps->vps_timing_info_present_flag = 1;
79
80 if (ctx->num_ticks_poc_diff_one > 0) {
81 vps->vps_num_ticks_poc_diff_one_minus1 =
82 ctx->num_ticks_poc_diff_one - 1;
83 vps->vps_poc_proportional_to_timing_flag = 1;
84 } else if (ctx->num_ticks_poc_diff_one == 0) {
85 vps->vps_poc_proportional_to_timing_flag = 0;
86 }
87 }
88
89 return 0;
90 }
91
92 static int h265_metadata_update_sps(AVBSFContext *bsf,
93 H265RawSPS *sps)
94 {
95 H265MetadataContext *ctx = bsf->priv_data;
96 int need_vui = 0;
97 int crop_unit_x, crop_unit_y;
98
99 if (ctx->sample_aspect_ratio.num && ctx->sample_aspect_ratio.den) {
100 // Table E-1.
101 static const AVRational sar_idc[] = {
102 { 0, 0 }, // Unspecified (never written here).
103 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
104 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
105 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
106 { 160, 99 }, { 4, 3 }, { 3, 2 }, { 2, 1 },
107 };
108 int num, den, i;
109
110 av_reduce(&num, &den, ctx->sample_aspect_ratio.num,
111 ctx->sample_aspect_ratio.den, 65535);
112
113 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
114 if (num == sar_idc[i].num &&
115 den == sar_idc[i].den)
116 break;
117 }
118 if (i == FF_ARRAY_ELEMS(sar_idc)) {
119 sps->vui.aspect_ratio_idc = 255;
120 sps->vui.sar_width = num;
121 sps->vui.sar_height = den;
122 } else {
123 sps->vui.aspect_ratio_idc = i;
124 }
125 sps->vui.aspect_ratio_info_present_flag = 1;
126 need_vui = 1;
127 }
128
129 #define SET_OR_INFER(field, value, present_flag, infer) do { \
130 if (value >= 0) { \
131 field = value; \
132 need_vui = 1; \
133 } else if (!present_flag) \
134 field = infer; \
135 } while (0)
136
137 if (ctx->video_format >= 0 ||
138 ctx->video_full_range_flag >= 0 ||
139 ctx->colour_primaries >= 0 ||
140 ctx->transfer_characteristics >= 0 ||
141 ctx->matrix_coefficients >= 0) {
142
143 SET_OR_INFER(sps->vui.video_format, ctx->video_format,
144 sps->vui.video_signal_type_present_flag, 5);
145
146 SET_OR_INFER(sps->vui.video_full_range_flag,
147 ctx->video_full_range_flag,
148 sps->vui.video_signal_type_present_flag, 0);
149
150 if (ctx->colour_primaries >= 0 ||
151 ctx->transfer_characteristics >= 0 ||
152 ctx->matrix_coefficients >= 0) {
153
154 SET_OR_INFER(sps->vui.colour_primaries,
155 ctx->colour_primaries,
156 sps->vui.colour_description_present_flag, 2);
157
158 SET_OR_INFER(sps->vui.transfer_characteristics,
159 ctx->transfer_characteristics,
160 sps->vui.colour_description_present_flag, 2);
161
162 SET_OR_INFER(sps->vui.matrix_coefficients,
163 ctx->matrix_coefficients,
164 sps->vui.colour_description_present_flag, 2);
165
166 sps->vui.colour_description_present_flag = 1;
167 }
168 sps->vui.video_signal_type_present_flag = 1;
169 need_vui = 1;
170 }
171
172 if (ctx->chroma_sample_loc_type >= 0) {
173 sps->vui.chroma_sample_loc_type_top_field =
174 ctx->chroma_sample_loc_type;
175 sps->vui.chroma_sample_loc_type_bottom_field =
176 ctx->chroma_sample_loc_type;
177 sps->vui.chroma_loc_info_present_flag = 1;
178 need_vui = 1;
179 }
180
181 if (ctx->tick_rate.num && ctx->tick_rate.den) {
182 int num, den;
183
184 av_reduce(&num, &den, ctx->tick_rate.num, ctx->tick_rate.den,
185 UINT32_MAX > INT_MAX ? UINT32_MAX : INT_MAX);
186
187 sps->vui.vui_time_scale = num;
188 sps->vui.vui_num_units_in_tick = den;
189
190 sps->vui.vui_timing_info_present_flag = 1;
191 need_vui = 1;
192
193 if (ctx->num_ticks_poc_diff_one > 0) {
194 sps->vui.vui_num_ticks_poc_diff_one_minus1 =
195 ctx->num_ticks_poc_diff_one - 1;
196 sps->vui.vui_poc_proportional_to_timing_flag = 1;
197 } else if (ctx->num_ticks_poc_diff_one == 0) {
198 sps->vui.vui_poc_proportional_to_timing_flag = 0;
199 }
200 }
201
202 if (sps->separate_colour_plane_flag || sps->chroma_format_idc == 0) {
203 crop_unit_x = 1;
204 crop_unit_y = 1;
205 } else {
206 crop_unit_x = 1 + (sps->chroma_format_idc < 3);
207 crop_unit_y = 1 + (sps->chroma_format_idc < 2);
208 }
209 #define CROP(border, unit) do { \
210 if (ctx->crop_ ## border >= 0) { \
211 if (ctx->crop_ ## border % unit != 0) { \
212 av_log(bsf, AV_LOG_ERROR, "Invalid value for crop_%s: " \
213 "must be a multiple of %d.\n", #border, unit); \
214 return AVERROR(EINVAL); \
215 } \
216 sps->conf_win_ ## border ## _offset = \
217 ctx->crop_ ## border / unit; \
218 sps->conformance_window_flag = 1; \
219 } \
220 } while (0)
221 CROP(left, crop_unit_x);
222 CROP(right, crop_unit_x);
223 CROP(top, crop_unit_y);
224 CROP(bottom, crop_unit_y);
225 #undef CROP
226
227 if (need_vui)
228 sps->vui_parameters_present_flag = 1;
229
230 return 0;
231 }
232
233 static int h265_metadata_filter(AVBSFContext *bsf, AVPacket *out)
234 {
235 H265MetadataContext *ctx = bsf->priv_data;
236 AVPacket *in = NULL;
237 CodedBitstreamFragment *au = &ctx->access_unit;
238 int err, i;
239
240 err = ff_bsf_get_packet(bsf, &in);
241 if (err < 0)
242 goto fail;
243
244 err = ff_cbs_read_packet(&ctx->cbc, au, in);
245 if (err < 0) {
246 av_log(bsf, AV_LOG_ERROR, "Failed to read packet.\n");
247 goto fail;
248 }
249
250 if (au->nb_units == 0) {
251 av_log(bsf, AV_LOG_ERROR, "No NAL units in packet.\n");
252 err = AVERROR_INVALIDDATA;
253 goto fail;
254 }
255
256 // If an AUD is present, it must be the first NAL unit.
257 if (au->units[0].type == HEVC_NAL_AUD) {
258 if (ctx->aud == REMOVE)
259 ff_cbs_delete_unit(&ctx->cbc, au, 0);
260 } else {
261 if (ctx->aud == INSERT) {
262 H265RawAUD *aud = &ctx->aud_nal;
263 int pic_type = 0, temporal_id = 8, layer_id = 0;
264
265 for (i = 0; i < au->nb_units; i++) {
266 const H265RawNALUnitHeader *nal = au->units[i].content;
267 if (!nal)
268 continue;
269 if (nal->nuh_temporal_id_plus1 < temporal_id + 1)
270 temporal_id = nal->nuh_temporal_id_plus1 - 1;
271
272 if (au->units[i].type <= HEVC_NAL_RSV_VCL31) {
273 const H265RawSlice *slice = au->units[i].content;
274 layer_id = nal->nuh_layer_id;
275 if (slice->header.slice_type == HEVC_SLICE_B &&
276 pic_type < 2)
277 pic_type = 2;
278 if (slice->header.slice_type == HEVC_SLICE_P &&
279 pic_type < 1)
280 pic_type = 1;
281 }
282 }
283
284 aud->nal_unit_header = (H265RawNALUnitHeader) {
285 .nal_unit_type = HEVC_NAL_AUD,
286 .nuh_layer_id = layer_id,
287 .nuh_temporal_id_plus1 = temporal_id + 1,
288 };
289 aud->pic_type = pic_type;
290
291 err = ff_cbs_insert_unit_content(&ctx->cbc, au,
292 0, HEVC_NAL_AUD, aud);
293 if (err) {
294 av_log(bsf, AV_LOG_ERROR, "Failed to insert AUD.\n");
295 goto fail;
296 }
297 }
298 }
299
300 for (i = 0; i < au->nb_units; i++) {
301 if (au->units[i].type == HEVC_NAL_VPS) {
302 err = h265_metadata_update_vps(bsf, au->units[i].content);
303 if (err < 0)
304 goto fail;
305 }
306 if (au->units[i].type == HEVC_NAL_SPS) {
307 err = h265_metadata_update_sps(bsf, au->units[i].content);
308 if (err < 0)
309 goto fail;
310 }
311 }
312
313 err = ff_cbs_write_packet(&ctx->cbc, out, au);
314 if (err < 0) {
315 av_log(bsf, AV_LOG_ERROR, "Failed to write packet.\n");
316 goto fail;
317 }
318
319 err = av_packet_copy_props(out, in);
320 if (err < 0)
321 goto fail;
322
323 err = 0;
324 fail:
325 ff_cbs_fragment_uninit(&ctx->cbc, au);
326
327 av_packet_free(&in);
328
329 return err;
330 }
331
332 static int h265_metadata_init(AVBSFContext *bsf)
333 {
334 H265MetadataContext *ctx = bsf->priv_data;
335 CodedBitstreamFragment *au = &ctx->access_unit;
336 int err, i;
337
338 err = ff_cbs_init(&ctx->cbc, AV_CODEC_ID_HEVC, bsf);
339 if (err < 0)
340 return err;
341
342 if (bsf->par_in->extradata) {
343 err = ff_cbs_read_extradata(&ctx->cbc, au, bsf->par_in);
344 if (err < 0) {
345 av_log(bsf, AV_LOG_ERROR, "Failed to read extradata.\n");
346 goto fail;
347 }
348
349 for (i = 0; i < au->nb_units; i++) {
350 if (au->units[i].type == HEVC_NAL_VPS) {
351 err = h265_metadata_update_vps(bsf, au->units[i].content);
352 if (err < 0)
353 goto fail;
354 }
355 if (au->units[i].type == HEVC_NAL_SPS) {
356 err = h265_metadata_update_sps(bsf, au->units[i].content);
357 if (err < 0)
358 goto fail;
359 }
360 }
361
362 err = ff_cbs_write_extradata(&ctx->cbc, bsf->par_out, au);
363 if (err < 0) {
364 av_log(bsf, AV_LOG_ERROR, "Failed to write extradata.\n");
365 goto fail;
366 }
367 }
368
369 err = 0;
370 fail:
371 ff_cbs_fragment_uninit(&ctx->cbc, au);
372 return err;
373 }
374
375 static void h265_metadata_close(AVBSFContext *bsf)
376 {
377 H265MetadataContext *ctx = bsf->priv_data;
378 ff_cbs_close(&ctx->cbc);
379 }
380
381 #define OFFSET(x) offsetof(H265MetadataContext, x)
382 static const AVOption h265_metadata_options[] = {
383 { "aud", "Access Unit Delimiter NAL units",
384 OFFSET(aud), AV_OPT_TYPE_INT,
385 { .i64 = PASS }, PASS, REMOVE, 0, "aud" },
386 { "pass", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PASS }, .unit = "aud" },
387 { "insert", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = INSERT }, .unit = "aud" },
388 { "remove", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = REMOVE }, .unit = "aud" },
389
390 { "sample_aspect_ratio", "Set sample aspect ratio (table E-1)",
391 OFFSET(sample_aspect_ratio), AV_OPT_TYPE_RATIONAL,
392 { .i64 = 0 }, 0, 65535 },
393
394 { "video_format", "Set video format (table E-2)",
395 OFFSET(video_format), AV_OPT_TYPE_INT,
396 { .i64 = -1 }, -1, 7 },
397 { "video_full_range_flag", "Set video full range flag",
398 OFFSET(video_full_range_flag), AV_OPT_TYPE_INT,
399 { .i64 = -1 }, -1, 1 },
400 { "colour_primaries", "Set colour primaries (table E-3)",
401 OFFSET(colour_primaries), AV_OPT_TYPE_INT,
402 { .i64 = -1 }, -1, 255 },
403 { "transfer_characteristics", "Set transfer characteristics (table E-4)",
404 OFFSET(transfer_characteristics), AV_OPT_TYPE_INT,
405 { .i64 = -1 }, -1, 255 },
406 { "matrix_coefficients", "Set matrix coefficients (table E-5)",
407 OFFSET(matrix_coefficients), AV_OPT_TYPE_INT,
408 { .i64 = -1 }, -1, 255 },
409
410 { "chroma_sample_loc_type", "Set chroma sample location type (figure E-1)",
411 OFFSET(chroma_sample_loc_type), AV_OPT_TYPE_INT,
412 { .i64 = -1 }, -1, 6 },
413
414 { "tick_rate",
415 "Set VPS and VUI tick rate (num_units_in_tick / time_scale)",
416 OFFSET(tick_rate), AV_OPT_TYPE_RATIONAL,
417 { .i64 = 0 }, 0, UINT_MAX },
418 { "num_ticks_poc_diff_one",
419 "Set VPS and VUI number of ticks per POC increment",
420 OFFSET(num_ticks_poc_diff_one), AV_OPT_TYPE_INT,
421 { .i64 = -1 }, -1, INT_MAX },
422
423 { "crop_left", "Set left border crop offset",
424 OFFSET(crop_left), AV_OPT_TYPE_INT,
425 { .i64 = -1 }, -1, HEVC_MAX_WIDTH },
426 { "crop_right", "Set right border crop offset",
427 OFFSET(crop_right), AV_OPT_TYPE_INT,
428 { .i64 = -1 }, -1, HEVC_MAX_WIDTH },
429 { "crop_top", "Set top border crop offset",
430 OFFSET(crop_top), AV_OPT_TYPE_INT,
431 { .i64 = -1 }, -1, HEVC_MAX_HEIGHT },
432 { "crop_bottom", "Set bottom border crop offset",
433 OFFSET(crop_bottom), AV_OPT_TYPE_INT,
434 { .i64 = -1 }, -1, HEVC_MAX_HEIGHT },
435
436 { NULL }
437 };
438
439 static const AVClass h265_metadata_class = {
440 .class_name = "h265_metadata_bsf",
441 .item_name = av_default_item_name,
442 .option = h265_metadata_options,
443 .version = LIBAVCODEC_VERSION_MAJOR,
444 };
445
446 static const enum AVCodecID h265_metadata_codec_ids[] = {
447 AV_CODEC_ID_HEVC, AV_CODEC_ID_NONE,
448 };
449
450 const AVBitStreamFilter ff_hevc_metadata_bsf = {
451 .name = "hevc_metadata",
452 .priv_data_size = sizeof(H265MetadataContext),
453 .priv_class = &h265_metadata_class,
454 .init = &h265_metadata_init,
455 .close = &h265_metadata_close,
456 .filter = &h265_metadata_filter,
457 .codec_ids = h265_metadata_codec_ids,
458 };