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/avstring.h"
20 #include "libavutil/display.h"
21 #include "libavutil/common.h"
22 #include "libavutil/opt.h"
42 typedef struct H264MetadataContext {
45 CodedBitstreamContext *cbc;
46 CodedBitstreamFragment access_unit;
52 AVRational sample_aspect_ratio;
55 int video_full_range_flag;
57 int transfer_characteristics;
58 int matrix_coefficients;
60 int chroma_sample_loc_type;
63 int fixed_frame_rate_flag;
70 const char *sei_user_data;
74 int display_orientation;
77 } H264MetadataContext;
80 static int h264_metadata_update_sps(AVBSFContext *bsf,
83 H264MetadataContext *ctx = bsf->priv_data;
85 int crop_unit_x, crop_unit_y;
87 if (ctx->sample_aspect_ratio.num && ctx->sample_aspect_ratio.den) {
89 static const AVRational sar_idc[] = {
90 { 0, 0 }, // Unspecified (never written here).
91 { 1, 1 }, { 12, 11 }, { 10, 11 }, { 16, 11 },
92 { 40, 33 }, { 24, 11 }, { 20, 11 }, { 32, 11 },
93 { 80, 33 }, { 18, 11 }, { 15, 11 }, { 64, 33 },
94 { 160, 99 }, { 4, 3 }, { 3, 2 }, { 2, 1 },
98 av_reduce(&num, &den, ctx->sample_aspect_ratio.num,
99 ctx->sample_aspect_ratio.den, 65535);
101 for (i = 1; i < FF_ARRAY_ELEMS(sar_idc); i++) {
102 if (num == sar_idc[i].num &&
103 den == sar_idc[i].den)
106 if (i == FF_ARRAY_ELEMS(sar_idc)) {
107 sps->vui.aspect_ratio_idc = 255;
108 sps->vui.sar_width = num;
109 sps->vui.sar_height = den;
111 sps->vui.aspect_ratio_idc = i;
113 sps->vui.aspect_ratio_info_present_flag = 1;
117 #define SET_OR_INFER(field, value, present_flag, infer) do { \
121 } else if (!present_flag) \
125 if (ctx->video_format >= 0 ||
126 ctx->video_full_range_flag >= 0 ||
127 ctx->colour_primaries >= 0 ||
128 ctx->transfer_characteristics >= 0 ||
129 ctx->matrix_coefficients >= 0) {
131 SET_OR_INFER(sps->vui.video_format, ctx->video_format,
132 sps->vui.video_signal_type_present_flag, 5);
134 SET_OR_INFER(sps->vui.video_full_range_flag,
135 ctx->video_full_range_flag,
136 sps->vui.video_signal_type_present_flag, 0);
138 if (ctx->colour_primaries >= 0 ||
139 ctx->transfer_characteristics >= 0 ||
140 ctx->matrix_coefficients >= 0) {
142 SET_OR_INFER(sps->vui.colour_primaries,
143 ctx->colour_primaries,
144 sps->vui.colour_description_present_flag, 2);
146 SET_OR_INFER(sps->vui.transfer_characteristics,
147 ctx->transfer_characteristics,
148 sps->vui.colour_description_present_flag, 2);
150 SET_OR_INFER(sps->vui.matrix_coefficients,
151 ctx->matrix_coefficients,
152 sps->vui.colour_description_present_flag, 2);
154 sps->vui.colour_description_present_flag = 1;
156 sps->vui.video_signal_type_present_flag = 1;
160 if (ctx->chroma_sample_loc_type >= 0) {
161 sps->vui.chroma_sample_loc_type_top_field =
162 ctx->chroma_sample_loc_type;
163 sps->vui.chroma_sample_loc_type_bottom_field =
164 ctx->chroma_sample_loc_type;
165 sps->vui.chroma_loc_info_present_flag = 1;
169 if (ctx->tick_rate.num && ctx->tick_rate.den) {
172 av_reduce(&num, &den, ctx->tick_rate.num, ctx->tick_rate.den,
173 UINT32_MAX > INT_MAX ? UINT32_MAX : INT_MAX);
175 sps->vui.time_scale = num;
176 sps->vui.num_units_in_tick = den;
178 sps->vui.timing_info_present_flag = 1;
181 SET_OR_INFER(sps->vui.fixed_frame_rate_flag,
182 ctx->fixed_frame_rate_flag,
183 sps->vui.timing_info_present_flag, 0);
185 if (sps->separate_colour_plane_flag || sps->chroma_format_idc == 0) {
187 crop_unit_y = 2 - sps->frame_mbs_only_flag;
189 crop_unit_x = 1 + (sps->chroma_format_idc < 3);
190 crop_unit_y = (1 + (sps->chroma_format_idc < 2)) *
191 (2 - sps->frame_mbs_only_flag);
193 #define CROP(border, unit) do { \
194 if (ctx->crop_ ## border >= 0) { \
195 if (ctx->crop_ ## border % unit != 0) { \
196 av_log(bsf, AV_LOG_ERROR, "Invalid value for crop_%s: " \
197 "must be a multiple of %d.\n", #border, unit); \
198 return AVERROR(EINVAL); \
200 sps->frame_crop_ ## border ## _offset = \
201 ctx->crop_ ## border / unit; \
202 sps->frame_cropping_flag = 1; \
205 CROP(left, crop_unit_x);
206 CROP(right, crop_unit_x);
207 CROP(top, crop_unit_y);
208 CROP(bottom, crop_unit_y);
212 sps->vui_parameters_present_flag = 1;
217 static int h264_metadata_filter(AVBSFContext *bsf, AVPacket *out)
219 H264MetadataContext *ctx = bsf->priv_data;
221 CodedBitstreamFragment *au = &ctx->access_unit;
222 int err, i, j, has_sps;
223 uint8_t *displaymatrix_side_data = NULL;
224 size_t displaymatrix_side_data_size = 0;
226 err = ff_bsf_get_packet(bsf, &in);
230 err = ff_cbs_read_packet(ctx->cbc, au, in);
232 av_log(bsf, AV_LOG_ERROR, "Failed to read packet.\n");
236 if (au->nb_units == 0) {
237 av_log(bsf, AV_LOG_ERROR, "No NAL units in packet.\n");
238 err = AVERROR_INVALIDDATA;
242 // If an AUD is present, it must be the first NAL unit.
243 if (au->units[0].type == H264_NAL_AUD) {
244 if (ctx->aud == REMOVE)
245 ff_cbs_delete_unit(ctx->cbc, au, 0);
247 if (ctx->aud == INSERT) {
248 static const int primary_pic_type_table[] = {
251 0x0e7, // 0, 1, 2, 5, 6, 7
255 0x3bd, // 0, 2, 3, 4, 5, 7, 8, 9
256 0x3ff, // 0, 1, 2, 3, 4, 5, 6, 7, 8, 9
258 int primary_pic_type_mask = 0xff;
260 .nal_unit_header.nal_unit_type = H264_NAL_AUD,
263 for (i = 0; i < au->nb_units; i++) {
264 if (au->units[i].type == H264_NAL_SLICE ||
265 au->units[i].type == H264_NAL_IDR_SLICE) {
266 H264RawSlice *slice = au->units[i].content;
267 for (j = 0; j < FF_ARRAY_ELEMS(primary_pic_type_table); j++) {
268 if (!(primary_pic_type_table[j] &
269 (1 << slice->header.slice_type)))
270 primary_pic_type_mask &= ~(1 << j);
274 for (j = 0; j < FF_ARRAY_ELEMS(primary_pic_type_table); j++)
275 if (primary_pic_type_mask & (1 << j))
277 if (j >= FF_ARRAY_ELEMS(primary_pic_type_table)) {
278 av_log(bsf, AV_LOG_ERROR, "No usable primary_pic_type: "
279 "invalid slice types?\n");
280 err = AVERROR_INVALIDDATA;
284 aud.primary_pic_type = j;
286 err = ff_cbs_insert_unit_content(ctx->cbc, au,
287 0, H264_NAL_AUD, &aud, NULL);
289 av_log(bsf, AV_LOG_ERROR, "Failed to insert AUD.\n");
296 for (i = 0; i < au->nb_units; i++) {
297 if (au->units[i].type == H264_NAL_SPS) {
298 err = h264_metadata_update_sps(bsf, au->units[i].content);
305 // Only insert the SEI in access units containing SPSs, and also
306 // unconditionally in the first access unit we ever see.
307 if (ctx->sei_user_data && (has_sps || !ctx->done_first_au)) {
308 H264RawSEIPayload payload = {
309 .payload_type = H264_SEI_TYPE_USER_DATA_UNREGISTERED,
311 H264RawSEIUserDataUnregistered *udu =
312 &payload.payload.user_data_unregistered;
314 for (i = j = 0; j < 32 && ctx->sei_user_data[i]; i++) {
316 c = ctx->sei_user_data[i];
319 } else if (av_isxdigit(c)) {
321 v = (c <= '9' ? c - '0' : c - 'a' + 10);
323 goto invalid_user_data;
326 udu->uuid_iso_iec_11578[j / 2] |= v;
328 udu->uuid_iso_iec_11578[j / 2] = v << 4;
331 if (j == 32 && ctx->sei_user_data[i] == '+') {
332 size_t len = strlen(ctx->sei_user_data + i + 1);
334 udu->data_ref = av_buffer_alloc(len + 1);
335 if (!udu->data_ref) {
336 err = AVERROR(ENOMEM);
340 udu->data = udu->data_ref->data;
341 udu->data_length = len + 1;
342 memcpy(udu->data, ctx->sei_user_data + i + 1, len + 1);
344 payload.payload_size = 16 + udu->data_length;
346 err = ff_cbs_h264_add_sei_message(ctx->cbc, au, &payload);
348 av_log(bsf, AV_LOG_ERROR, "Failed to add user data SEI "
349 "message to access unit.\n");
355 av_log(bsf, AV_LOG_ERROR, "Invalid user data: "
356 "must be \"UUID+string\".\n");
357 err = AVERROR(EINVAL);
362 if (ctx->delete_filler) {
363 for (i = 0; i < au->nb_units; i++) {
364 if (au->units[i].type == H264_NAL_FILLER_DATA) {
366 err = ff_cbs_delete_unit(ctx->cbc, au, i);
368 av_log(bsf, AV_LOG_ERROR, "Failed to delete "
376 if (au->units[i].type == H264_NAL_SEI) {
377 // Filler SEI messages.
378 H264RawSEI *sei = au->units[i].content;
380 for (j = 0; j < sei->payload_count; j++) {
381 if (sei->payload[j].payload_type ==
382 H264_SEI_TYPE_FILLER_PAYLOAD) {
383 err = ff_cbs_h264_delete_sei_message(ctx->cbc, au,
386 av_log(bsf, AV_LOG_ERROR, "Failed to delete "
387 "filler SEI message.\n");
390 // Renumbering might have happened, start again at
391 // the same NAL unit position.
400 if (ctx->display_orientation != PASS) {
401 for (i = 0; i < au->nb_units; i++) {
403 if (au->units[i].type != H264_NAL_SEI)
405 sei = au->units[i].content;
407 for (j = 0; j < sei->payload_count; j++) {
408 H264RawSEIDisplayOrientation *disp;
411 if (sei->payload[j].payload_type !=
412 H264_SEI_TYPE_DISPLAY_ORIENTATION)
414 disp = &sei->payload[j].payload.display_orientation;
416 if (ctx->display_orientation == REMOVE ||
417 ctx->display_orientation == INSERT) {
418 err = ff_cbs_h264_delete_sei_message(ctx->cbc, au,
421 av_log(bsf, AV_LOG_ERROR, "Failed to delete "
422 "display orientation SEI message.\n");
429 matrix = av_mallocz(9 * sizeof(int32_t));
431 err = AVERROR(ENOMEM);
435 av_display_rotation_set(matrix,
436 disp->anticlockwise_rotation *
438 av_display_matrix_flip(matrix, disp->hor_flip, disp->ver_flip);
440 // If there are multiple display orientation messages in an
441 // access unit then ignore all but the last one.
442 av_freep(&displaymatrix_side_data);
444 displaymatrix_side_data = (uint8_t*)matrix;
445 displaymatrix_side_data_size = 9 * sizeof(int32_t);
449 if (ctx->display_orientation == INSERT) {
450 H264RawSEIPayload payload = {
451 .payload_type = H264_SEI_TYPE_DISPLAY_ORIENTATION,
453 H264RawSEIDisplayOrientation *disp =
454 &payload.payload.display_orientation;
459 data = av_packet_get_side_data(in, AV_PKT_DATA_DISPLAYMATRIX, &size);
460 if (data && size >= 9 * sizeof(int32_t)) {
465 memcpy(matrix, data, sizeof(matrix));
468 if (matrix[0] < 0 && matrix[4] > 0)
470 else if (matrix[0] > 0 && matrix[4] < 0)
472 av_display_matrix_flip(matrix, hflip, vflip);
474 angle = av_display_rotation_get(matrix);
476 if (!(angle >= -180.0 && angle <= 180.0 /* also excludes NaN */) ||
477 matrix[2] != 0 || matrix[5] != 0 ||
478 matrix[6] != 0 || matrix[7] != 0) {
479 av_log(bsf, AV_LOG_WARNING, "Input display matrix is not "
480 "representable in H.264 parameters.\n");
482 disp->hor_flip = hflip;
483 disp->ver_flip = vflip;
484 disp->anticlockwise_rotation =
485 (uint16_t)rint((angle >= 0.0 ? angle
492 if (has_sps || !ctx->done_first_au) {
493 if (!isnan(ctx->rotate)) {
494 disp->anticlockwise_rotation =
495 (uint16_t)rint((ctx->rotate >= 0.0 ? ctx->rotate
496 : ctx->rotate + 360.0) *
501 disp->hor_flip = !!(ctx->flip & FLIP_HORIZONTAL);
502 disp->ver_flip = !!(ctx->flip & FLIP_VERTICAL);
508 disp->display_orientation_repetition_period = 1;
510 err = ff_cbs_h264_add_sei_message(ctx->cbc, au, &payload);
512 av_log(bsf, AV_LOG_ERROR, "Failed to add display orientation "
513 "SEI message to access unit.\n");
519 err = ff_cbs_write_packet(ctx->cbc, out, au);
521 av_log(bsf, AV_LOG_ERROR, "Failed to write packet.\n");
525 err = av_packet_copy_props(out, in);
529 if (displaymatrix_side_data) {
530 err = av_packet_add_side_data(out, AV_PKT_DATA_DISPLAYMATRIX,
531 displaymatrix_side_data,
532 displaymatrix_side_data_size);
534 av_log(bsf, AV_LOG_ERROR, "Failed to attach extracted "
535 "displaymatrix side data to packet.\n");
538 displaymatrix_side_data = NULL;
541 ctx->done_first_au = 1;
545 ff_cbs_fragment_uninit(ctx->cbc, au);
546 av_freep(&displaymatrix_side_data);
549 av_packet_unref(out);
555 static int h264_metadata_init(AVBSFContext *bsf)
557 H264MetadataContext *ctx = bsf->priv_data;
558 CodedBitstreamFragment *au = &ctx->access_unit;
561 err = ff_cbs_init(&ctx->cbc, AV_CODEC_ID_H264, bsf);
565 if (bsf->par_in->extradata) {
566 err = ff_cbs_read_extradata(ctx->cbc, au, bsf->par_in);
568 av_log(bsf, AV_LOG_ERROR, "Failed to read extradata.\n");
572 for (i = 0; i < au->nb_units; i++) {
573 if (au->units[i].type == H264_NAL_SPS) {
574 err = h264_metadata_update_sps(bsf, au->units[i].content);
580 err = ff_cbs_write_extradata(ctx->cbc, bsf->par_out, au);
582 av_log(bsf, AV_LOG_ERROR, "Failed to write extradata.\n");
589 ff_cbs_fragment_uninit(ctx->cbc, au);
593 static void h264_metadata_close(AVBSFContext *bsf)
595 H264MetadataContext *ctx = bsf->priv_data;
596 ff_cbs_close(&ctx->cbc);
599 #define OFFSET(x) offsetof(H264MetadataContext, x)
600 #define FLAGS (AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_BSF_PARAM)
601 static const AVOption h264_metadata_options[] = {
602 { "aud", "Access Unit Delimiter NAL units",
603 OFFSET(aud), AV_OPT_TYPE_INT,
604 { .i64 = PASS }, PASS, REMOVE, FLAGS, "aud" },
605 { "pass", NULL, 0, AV_OPT_TYPE_CONST,
606 { .i64 = PASS }, .flags = FLAGS, .unit = "aud" },
607 { "insert", NULL, 0, AV_OPT_TYPE_CONST,
608 { .i64 = INSERT }, .flags = FLAGS, .unit = "aud" },
609 { "remove", NULL, 0, AV_OPT_TYPE_CONST,
610 { .i64 = REMOVE }, .flags = FLAGS, .unit = "aud" },
612 { "sample_aspect_ratio", "Set sample aspect ratio (table E-1)",
613 OFFSET(sample_aspect_ratio), AV_OPT_TYPE_RATIONAL,
614 { .dbl = 0.0 }, 0, 65535, FLAGS },
616 { "video_format", "Set video format (table E-2)",
617 OFFSET(video_format), AV_OPT_TYPE_INT,
618 { .i64 = -1 }, -1, 7, FLAGS},
619 { "video_full_range_flag", "Set video full range flag",
620 OFFSET(video_full_range_flag), AV_OPT_TYPE_INT,
621 { .i64 = -1 }, -1, 1, FLAGS },
622 { "colour_primaries", "Set colour primaries (table E-3)",
623 OFFSET(colour_primaries), AV_OPT_TYPE_INT,
624 { .i64 = -1 }, -1, 255, FLAGS },
625 { "transfer_characteristics", "Set transfer characteristics (table E-4)",
626 OFFSET(transfer_characteristics), AV_OPT_TYPE_INT,
627 { .i64 = -1 }, -1, 255, FLAGS },
628 { "matrix_coefficients", "Set matrix coefficients (table E-5)",
629 OFFSET(matrix_coefficients), AV_OPT_TYPE_INT,
630 { .i64 = -1 }, -1, 255, FLAGS },
632 { "chroma_sample_loc_type", "Set chroma sample location type (figure E-1)",
633 OFFSET(chroma_sample_loc_type), AV_OPT_TYPE_INT,
634 { .i64 = -1 }, -1, 6, FLAGS },
636 { "tick_rate", "Set VUI tick rate (num_units_in_tick / time_scale)",
637 OFFSET(tick_rate), AV_OPT_TYPE_RATIONAL,
638 { .dbl = 0.0 }, 0, UINT_MAX, FLAGS },
639 { "fixed_frame_rate_flag", "Set VUI fixed frame rate flag",
640 OFFSET(fixed_frame_rate_flag), AV_OPT_TYPE_INT,
641 { .i64 = -1 }, -1, 1, FLAGS },
643 { "crop_left", "Set left border crop offset",
644 OFFSET(crop_left), AV_OPT_TYPE_INT,
645 { .i64 = -1 }, -1, H264_MAX_WIDTH, FLAGS },
646 { "crop_right", "Set right border crop offset",
647 OFFSET(crop_right), AV_OPT_TYPE_INT,
648 { .i64 = -1 }, -1, H264_MAX_WIDTH, FLAGS },
649 { "crop_top", "Set top border crop offset",
650 OFFSET(crop_top), AV_OPT_TYPE_INT,
651 { .i64 = -1 }, -1, H264_MAX_HEIGHT, FLAGS },
652 { "crop_bottom", "Set bottom border crop offset",
653 OFFSET(crop_bottom), AV_OPT_TYPE_INT,
654 { .i64 = -1 }, -1, H264_MAX_HEIGHT, FLAGS },
656 { "sei_user_data", "Insert SEI user data (UUID+string)",
657 OFFSET(sei_user_data), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
659 { "delete_filler", "Delete all filler (both NAL and SEI)",
660 OFFSET(delete_filler), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, FLAGS},
662 { "display_orientation", "Display orientation SEI",
663 OFFSET(display_orientation), AV_OPT_TYPE_INT,
664 { .i64 = PASS }, PASS, EXTRACT, FLAGS, "disp_or" },
665 { "pass", NULL, 0, AV_OPT_TYPE_CONST,
666 { .i64 = PASS }, .flags = FLAGS, .unit = "disp_or" },
667 { "insert", NULL, 0, AV_OPT_TYPE_CONST,
668 { .i64 = INSERT }, .flags = FLAGS, .unit = "disp_or" },
669 { "remove", NULL, 0, AV_OPT_TYPE_CONST,
670 { .i64 = REMOVE }, .flags = FLAGS, .unit = "disp_or" },
671 { "extract", NULL, 0, AV_OPT_TYPE_CONST,
672 { .i64 = EXTRACT }, .flags = FLAGS, .unit = "disp_or" },
674 { "rotate", "Set rotation in display orientation SEI (anticlockwise angle in degrees)",
675 OFFSET(rotate), AV_OPT_TYPE_DOUBLE,
676 { .dbl = NAN }, -360.0, +360.0, FLAGS },
677 { "flip", "Set flip in display orientation SEI",
678 OFFSET(flip), AV_OPT_TYPE_FLAGS,
679 { .i64 = 0 }, 0, FLIP_HORIZONTAL | FLIP_VERTICAL, FLAGS, "flip" },
680 { "horizontal", "Set hor_flip",
681 0, AV_OPT_TYPE_CONST,
682 { .i64 = FLIP_HORIZONTAL }, .flags = FLAGS, .unit = "flip" },
683 { "vertical", "Set ver_flip",
684 0, AV_OPT_TYPE_CONST,
685 { .i64 = FLIP_VERTICAL }, .flags = FLAGS, .unit = "flip" },
690 static const AVClass h264_metadata_class = {
691 .class_name = "h264_metadata_bsf",
692 .item_name = av_default_item_name,
693 .option = h264_metadata_options,
694 .version = LIBAVUTIL_VERSION_INT,
697 static const enum AVCodecID h264_metadata_codec_ids[] = {
698 AV_CODEC_ID_H264, AV_CODEC_ID_NONE,
701 const AVBitStreamFilter ff_h264_metadata_bsf = {
702 .name = "h264_metadata",
703 .priv_data_size = sizeof(H264MetadataContext),
704 .priv_class = &h264_metadata_class,
705 .init = &h264_metadata_init,
706 .close = &h264_metadata_close,
707 .filter = &h264_metadata_filter,
708 .codec_ids = h264_metadata_codec_ids,