avcodec/h264_sei: Remove "Subtitles with data type 0x%02x" sample request
[ffmpeg.git] / libavcodec / h264_sei.c
1 /*
2  * H.26L/H.264/AVC/JVT/14496-10/... sei decoding
3  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21
22 /**
23  * @file
24  * H.264 / AVC / MPEG4 part10 sei decoding.
25  * @author Michael Niedermayer <michaelni@gmx.at>
26  */
27
28 #include "avcodec.h"
29 #include "golomb.h"
30 #include "h264.h"
31 #include "internal.h"
32
33 static const uint8_t sei_num_clock_ts_table[9] = {
34     1, 1, 1, 2, 2, 3, 3, 2, 3
35 };
36
37 void ff_h264_reset_sei(H264Context *h)
38 {
39     h->sei_recovery_frame_cnt       = -1;
40     h->sei_dpb_output_delay         =  0;
41     h->sei_cpb_removal_delay        = -1;
42     h->sei_buffering_period_present =  0;
43     h->sei_frame_packing_present    =  0;
44     h->sei_display_orientation_present = 0;
45     h->sei_reguserdata_afd_present  =  0;
46
47     h->a53_caption_size = 0;
48     av_freep(&h->a53_caption);
49 }
50
51 static int decode_picture_timing(H264Context *h)
52 {
53     SPS *sps = &h->sps;
54     int i;
55
56     for (i = 0; i<MAX_SPS_COUNT; i++)
57         if (!sps->log2_max_frame_num && h->sps_buffers[i])
58             sps = h->sps_buffers[i];
59
60     if (sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag) {
61         h->sei_cpb_removal_delay = get_bits_long(&h->gb,
62                                                  sps->cpb_removal_delay_length);
63         h->sei_dpb_output_delay  = get_bits_long(&h->gb,
64                                                  sps->dpb_output_delay_length);
65     }
66     if (sps->pic_struct_present_flag) {
67         unsigned int i, num_clock_ts;
68
69         h->sei_pic_struct = get_bits(&h->gb, 4);
70         h->sei_ct_type    = 0;
71
72         if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
73             return AVERROR_INVALIDDATA;
74
75         num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
76
77         for (i = 0; i < num_clock_ts; i++) {
78             if (get_bits(&h->gb, 1)) {                /* clock_timestamp_flag */
79                 unsigned int full_timestamp_flag;
80
81                 h->sei_ct_type |= 1 << get_bits(&h->gb, 2);
82                 skip_bits(&h->gb, 1);                 /* nuit_field_based_flag */
83                 skip_bits(&h->gb, 5);                 /* counting_type */
84                 full_timestamp_flag = get_bits(&h->gb, 1);
85                 skip_bits(&h->gb, 1);                 /* discontinuity_flag */
86                 skip_bits(&h->gb, 1);                 /* cnt_dropped_flag */
87                 skip_bits(&h->gb, 8);                 /* n_frames */
88                 if (full_timestamp_flag) {
89                     skip_bits(&h->gb, 6);             /* seconds_value 0..59 */
90                     skip_bits(&h->gb, 6);             /* minutes_value 0..59 */
91                     skip_bits(&h->gb, 5);             /* hours_value 0..23 */
92                 } else {
93                     if (get_bits(&h->gb, 1)) {        /* seconds_flag */
94                         skip_bits(&h->gb, 6);         /* seconds_value range 0..59 */
95                         if (get_bits(&h->gb, 1)) {    /* minutes_flag */
96                             skip_bits(&h->gb, 6);     /* minutes_value 0..59 */
97                             if (get_bits(&h->gb, 1))  /* hours_flag */
98                                 skip_bits(&h->gb, 5); /* hours_value 0..23 */
99                         }
100                     }
101                 }
102                 if (sps->time_offset_length > 0)
103                     skip_bits(&h->gb,
104                               sps->time_offset_length); /* time_offset */
105             }
106         }
107
108         if (h->avctx->debug & FF_DEBUG_PICT_INFO)
109             av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n",
110                    h->sei_ct_type, h->sei_pic_struct);
111     }
112     return 0;
113 }
114
115 static int decode_registered_user_data_afd(H264Context *h, int size)
116 {
117     int flag;
118
119     if (size-- < 1)
120         return AVERROR_INVALIDDATA;
121     skip_bits(&h->gb, 1);               // 0
122     flag = get_bits(&h->gb, 1);         // active_format_flag
123     skip_bits(&h->gb, 6);               // reserved
124
125     if (flag) {
126         if (size-- < 1)
127             return AVERROR_INVALIDDATA;
128         skip_bits(&h->gb, 4);           // reserved
129         h->active_format_description   = get_bits(&h->gb, 4);
130         h->sei_reguserdata_afd_present = 1;
131 #if FF_API_AFD
132 FF_DISABLE_DEPRECATION_WARNINGS
133         h->avctx->dtg_active_format = h->active_format_description;
134 FF_ENABLE_DEPRECATION_WARNINGS
135 #endif /* FF_API_AFD */
136     }
137
138     return 0;
139 }
140
141 static int decode_registered_user_data_closed_caption(H264Context *h, int size)
142 {
143     int flag;
144     int user_data_type_code;
145     int cc_count;
146
147     if (size < 3)
148         return AVERROR(EINVAL);
149
150     user_data_type_code = get_bits(&h->gb, 8);
151     if (user_data_type_code == 0x3) {
152         skip_bits(&h->gb, 1);           // reserved
153
154         flag = get_bits(&h->gb, 1);     // process_cc_data_flag
155         if (flag) {
156             skip_bits(&h->gb, 1);       // zero bit
157             cc_count = get_bits(&h->gb, 5);
158             skip_bits(&h->gb, 8);       // reserved
159             size -= 2;
160
161             if (cc_count && size >= cc_count * 3) {
162                 const uint64_t new_size = (h->a53_caption_size + cc_count
163                                            * UINT64_C(3));
164                 int i, ret;
165
166                 if (new_size > INT_MAX)
167                     return AVERROR(EINVAL);
168
169                 /* Allow merging of the cc data from two fields. */
170                 ret = av_reallocp(&h->a53_caption, new_size);
171                 if (ret < 0)
172                     return ret;
173
174                 for (i = 0; i < cc_count; i++) {
175                     h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
176                     h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
177                     h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
178                 }
179
180                 skip_bits(&h->gb, 8);   // marker_bits
181             }
182         }
183     } else {
184         int i;
185         for (i = 0; i < size - 1; i++)
186             skip_bits(&h->gb, 8);
187     }
188
189     return 0;
190 }
191
192 static int decode_registered_user_data(H264Context *h, int size)
193 {
194     uint32_t country_code;
195     uint32_t user_identifier;
196
197     if (size < 7)
198         return AVERROR_INVALIDDATA;
199     size -= 7;
200
201     country_code = get_bits(&h->gb, 8); // itu_t_t35_country_code
202     if (country_code == 0xFF) {
203         skip_bits(&h->gb, 8);           // itu_t_t35_country_code_extension_byte
204         size--;
205     }
206
207     /* itu_t_t35_payload_byte follows */
208     skip_bits(&h->gb, 8);              // terminal provider code
209     skip_bits(&h->gb, 8);              // terminal provider oriented code
210     user_identifier = get_bits_long(&h->gb, 32);
211
212     switch (user_identifier) {
213         case MKBETAG('D', 'T', 'G', '1'):       // afd_data
214             return decode_registered_user_data_afd(h, size);
215         case MKBETAG('G', 'A', '9', '4'):       // closed captions
216             return decode_registered_user_data_closed_caption(h, size);
217         default:
218             skip_bits(&h->gb, size * 8);
219             break;
220     }
221
222     return 0;
223 }
224
225 static int decode_unregistered_user_data(H264Context *h, int size)
226 {
227     uint8_t user_data[16 + 256];
228     int e, build, i;
229
230     if (size < 16)
231         return AVERROR_INVALIDDATA;
232
233     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
234         user_data[i] = get_bits(&h->gb, 8);
235
236     user_data[i] = 0;
237     e = sscanf(user_data + 16, "x264 - core %d", &build);
238     if (e == 1 && build > 0)
239         h->x264_build = build;
240     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
241         h->x264_build = 67;
242
243     if (h->avctx->debug & FF_DEBUG_BUGS)
244         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
245
246     for (; i < size; i++)
247         skip_bits(&h->gb, 8);
248
249     return 0;
250 }
251
252 static int decode_recovery_point(H264Context *h)
253 {
254     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
255
256     /* 1b exact_match_flag,
257      * 1b broken_link_flag,
258      * 2b changing_slice_group_idc */
259     skip_bits(&h->gb, 4);
260
261     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
262         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
263
264     h->has_recovery_point = 1;
265
266     return 0;
267 }
268
269 static int decode_buffering_period(H264Context *h)
270 {
271     unsigned int sps_id;
272     int sched_sel_idx;
273     SPS *sps;
274
275     sps_id = get_ue_golomb_31(&h->gb);
276     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
277         av_log(h->avctx, AV_LOG_ERROR,
278                "non-existing SPS %d referenced in buffering period\n", sps_id);
279         return AVERROR_INVALIDDATA;
280     }
281     sps = h->sps_buffers[sps_id];
282
283     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
284     if (sps->nal_hrd_parameters_present_flag) {
285         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
286             h->initial_cpb_removal_delay[sched_sel_idx] =
287                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
288             // initial_cpb_removal_delay_offset
289             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
290         }
291     }
292     if (sps->vcl_hrd_parameters_present_flag) {
293         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
294             h->initial_cpb_removal_delay[sched_sel_idx] =
295                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
296             // initial_cpb_removal_delay_offset
297             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
298         }
299     }
300
301     h->sei_buffering_period_present = 1;
302     return 0;
303 }
304
305 static int decode_frame_packing_arrangement(H264Context *h)
306 {
307     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
308     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
309     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
310
311     if (h->sei_frame_packing_present) {
312         h->sei_fpa.frame_packing_arrangement_type =
313         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
314         h->sei_fpa.quincunx_sampling_flag         =
315         h->quincunx_subsampling           = get_bits1(&h->gb);
316         h->sei_fpa.content_interpretation_type    =
317         h->content_interpretation_type    = get_bits(&h->gb, 6);
318
319         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
320         // field_views_flag, current_frame_is_frame0_flag,
321         // frame0_self_contained_flag, frame1_self_contained_flag
322         skip_bits(&h->gb, 6);
323
324         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
325             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
326         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
327         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
328     }
329     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
330
331     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
332         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
333                                        h->sei_fpa.frame_packing_arrangement_id,
334                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
335                                        h->sei_fpa.frame_packing_arrangement_type,
336                                        h->sei_fpa.quincunx_sampling_flag,
337                                        h->sei_fpa.content_interpretation_type,
338                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
339
340     return 0;
341 }
342
343 static int decode_display_orientation(H264Context *h)
344 {
345     h->sei_display_orientation_present = !get_bits1(&h->gb);
346
347     if (h->sei_display_orientation_present) {
348         h->sei_hflip = get_bits1(&h->gb);     // hor_flip
349         h->sei_vflip = get_bits1(&h->gb);     // ver_flip
350
351         h->sei_anticlockwise_rotation = get_bits(&h->gb, 16);
352         get_ue_golomb(&h->gb);  // display_orientation_repetition_period
353         skip_bits1(&h->gb);     // display_orientation_extension_flag
354     }
355
356     return 0;
357 }
358
359 static int decode_GreenMetadata(H264Context *h)
360 {
361     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
362         av_log(h->avctx, AV_LOG_DEBUG,          "Green Metadata Info SEI message\n");
363
364     h->sei_green_metadata.green_metadata_type=get_bits(&h->gb, 8);
365
366     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
367         av_log(h->avctx, AV_LOG_DEBUG,          "green_metadata_type                            = %d\n",
368                h->sei_green_metadata.green_metadata_type);
369
370     if (h->sei_green_metadata.green_metadata_type==0){
371         h->sei_green_metadata.period_type=get_bits(&h->gb, 8);
372
373         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
374             av_log(h->avctx, AV_LOG_DEBUG,      "green_metadata_period_type                     = %d\n",
375                    h->sei_green_metadata.period_type);
376
377         if (h->sei_green_metadata.green_metadata_type==2){
378             h->sei_green_metadata.num_seconds = get_bits(&h->gb, 16);
379             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
380                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_seconds                     = %d\n",
381                        h->sei_green_metadata.num_seconds);
382         }
383         else if (h->sei_green_metadata.period_type==3){
384             h->sei_green_metadata.num_pictures = get_bits(&h->gb, 16);
385             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
386                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_pictures                    = %d\n",
387                        h->sei_green_metadata.num_pictures);
388         }
389
390         h->sei_green_metadata.percent_non_zero_macroblocks=get_bits(&h->gb, 8);
391         h->sei_green_metadata.percent_intra_coded_macroblocks=get_bits(&h->gb, 8);
392         h->sei_green_metadata.percent_six_tap_filtering=get_bits(&h->gb, 8);
393         h->sei_green_metadata.percent_alpha_point_deblocking_instance=get_bits(&h->gb, 8);
394
395         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
396             av_log(h->avctx, AV_LOG_DEBUG,      "SEI GREEN Complexity Metrics                   = %f %f %f %f\n",
397                                            (float)h->sei_green_metadata.percent_non_zero_macroblocks/255,
398                                            (float)h->sei_green_metadata.percent_intra_coded_macroblocks/255,
399                                            (float)h->sei_green_metadata.percent_six_tap_filtering/255,
400                                            (float)h->sei_green_metadata.percent_alpha_point_deblocking_instance/255);
401
402     }else if( h->sei_green_metadata.green_metadata_type==1){
403         h->sei_green_metadata.xsd_metric_type=get_bits(&h->gb, 8);
404         h->sei_green_metadata.xsd_metric_value=get_bits(&h->gb, 16);
405
406         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
407             av_log(h->avctx, AV_LOG_DEBUG,      "xsd_metric_type                                = %d\n",
408                    h->sei_green_metadata.xsd_metric_type);
409         if ( h->sei_green_metadata.xsd_metric_type==0){
410             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
411                 av_log(h->avctx, AV_LOG_DEBUG,  "xsd_metric_value                               = %f\n",
412                        (float)h->sei_green_metadata.xsd_metric_value/100);
413         }
414     }
415
416     return 0;
417 }
418
419 int ff_h264_decode_sei(H264Context *h)
420 {
421     while (get_bits_left(&h->gb) > 16 && show_bits(&h->gb, 16)) {
422         int type = 0;
423         unsigned size = 0;
424         unsigned next;
425         int ret  = 0;
426
427         do {
428             if (get_bits_left(&h->gb) < 8)
429                 return AVERROR_INVALIDDATA;
430             type += show_bits(&h->gb, 8);
431         } while (get_bits(&h->gb, 8) == 255);
432
433         do {
434             if (get_bits_left(&h->gb) < 8)
435                 return AVERROR_INVALIDDATA;
436             size += show_bits(&h->gb, 8);
437         } while (get_bits(&h->gb, 8) == 255);
438
439         if (h->avctx->debug&FF_DEBUG_STARTCODE)
440             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
441
442         if (size > get_bits_left(&h->gb) / 8) {
443             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
444                    type, 8*size, get_bits_left(&h->gb));
445             return AVERROR_INVALIDDATA;
446         }
447         next = get_bits_count(&h->gb) + 8 * size;
448
449         switch (type) {
450         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
451             ret = decode_picture_timing(h);
452             break;
453         case SEI_TYPE_USER_DATA_REGISTERED:
454             ret = decode_registered_user_data(h, size);
455             break;
456         case SEI_TYPE_USER_DATA_UNREGISTERED:
457             ret = decode_unregistered_user_data(h, size);
458             break;
459         case SEI_TYPE_RECOVERY_POINT:
460             ret = decode_recovery_point(h);
461             break;
462         case SEI_TYPE_BUFFERING_PERIOD:
463             ret = decode_buffering_period(h);
464             break;
465         case SEI_TYPE_FRAME_PACKING:
466             ret = decode_frame_packing_arrangement(h);
467             break;
468         case SEI_TYPE_DISPLAY_ORIENTATION:
469             ret = decode_display_orientation(h);
470             break;
471         case SEI_TYPE_GREEN_METADATA:
472             ret = decode_GreenMetadata(h);
473             break;
474         default:
475             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
476         }
477         if (ret < 0)
478             return ret;
479
480         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
481
482         // FIXME check bits here
483         align_get_bits(&h->gb);
484     }
485
486     return 0;
487 }
488
489 const char* ff_h264_sei_stereo_mode(H264Context *h)
490 {
491     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
492         switch (h->sei_fpa.frame_packing_arrangement_type) {
493             case SEI_FPA_TYPE_CHECKERBOARD:
494                 if (h->sei_fpa.content_interpretation_type == 2)
495                     return "checkerboard_rl";
496                 else
497                     return "checkerboard_lr";
498             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
499                 if (h->sei_fpa.content_interpretation_type == 2)
500                     return "col_interleaved_rl";
501                 else
502                     return "col_interleaved_lr";
503             case SEI_FPA_TYPE_INTERLEAVE_ROW:
504                 if (h->sei_fpa.content_interpretation_type == 2)
505                     return "row_interleaved_rl";
506                 else
507                     return "row_interleaved_lr";
508             case SEI_FPA_TYPE_SIDE_BY_SIDE:
509                 if (h->sei_fpa.content_interpretation_type == 2)
510                     return "right_left";
511                 else
512                     return "left_right";
513             case SEI_FPA_TYPE_TOP_BOTTOM:
514                 if (h->sei_fpa.content_interpretation_type == 2)
515                     return "bottom_top";
516                 else
517                     return "top_bottom";
518             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
519                 if (h->sei_fpa.content_interpretation_type == 2)
520                     return "block_rl";
521                 else
522                     return "block_lr";
523             case SEI_FPA_TYPE_2D:
524             default:
525                 return "mono";
526         }
527     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
528         return "mono";
529     } else {
530         return NULL;
531     }
532 }