avcodec: Add support for Closed Caption export in h264
[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     if (h->a53_caption)
47         av_freep(&h->a53_caption);
48     h->a53_caption_size = 0;
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(H264Context *h, int size)
116 {
117     uint32_t country_code;
118     uint32_t user_identifier;
119     int flag, cc_count, user_data_type_code;
120
121     if (size < 7)
122         return AVERROR_INVALIDDATA;
123     size -= 7;
124
125     country_code = get_bits(&h->gb, 8); // itu_t_t35_country_code
126     if (country_code == 0xFF) {
127         skip_bits(&h->gb, 8);           // itu_t_t35_country_code_extension_byte
128         size--;
129     }
130
131     /* itu_t_t35_payload_byte follows */
132     skip_bits(&h->gb, 8);              // terminal provider code
133     skip_bits(&h->gb, 8);              // terminal provider oriented code
134     user_identifier = get_bits_long(&h->gb, 32);
135
136     switch (user_identifier) {
137         case MKBETAG('D', 'T', 'G', '1'):       // afd_data
138             if (size-- < 1)
139                 return AVERROR_INVALIDDATA;
140             skip_bits(&h->gb, 1);               // 0
141             flag = get_bits(&h->gb, 1);         // active_format_flag
142             skip_bits(&h->gb, 6);               // reserved
143
144             if (flag) {
145                 if (size-- < 1)
146                     return AVERROR_INVALIDDATA;
147                 skip_bits(&h->gb, 4);           // reserved
148                 h->active_format_description   = get_bits(&h->gb, 4);
149                 h->sei_reguserdata_afd_present = 1;
150 #if FF_API_AFD
151 FF_DISABLE_DEPRECATION_WARNINGS
152                 h->avctx->dtg_active_format = h->active_format_description;
153 FF_ENABLE_DEPRECATION_WARNINGS
154 #endif /* FF_API_AFD */
155             }
156             break;
157         case MKBETAG('G', 'A', '9', '4'): // "GA94" closed captions
158             if (size < 3)
159                 return AVERROR(EINVAL);
160             user_data_type_code = get_bits(&h->gb, 8);
161             if (user_data_type_code == 0x3) {
162                 skip_bits(&h->gb, 1);
163                 if (get_bits(&h->gb, 1)) {
164                     skip_bits(&h->gb, 1);
165                     cc_count = get_bits(&h->gb, 5);
166                     skip_bits(&h->gb, 8);
167                     size -= 2;
168                     if (cc_count && size >= cc_count*3) {
169                         int i;
170                         uint8_t *tmp;
171                         if ((int64_t)h->a53_caption_size + (int64_t)cc_count*3 > INT_MAX)
172                             return AVERROR(EINVAL);
173
174                         // Allow merging of the cc data from two fields
175                         tmp = av_realloc(h->a53_caption, h->a53_caption_size + cc_count*3);
176                         if (!tmp)
177                             return AVERROR(ENOMEM);
178                         h->a53_caption = tmp;
179                         for (i = 0; i < cc_count; i++) {
180                             h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
181                             h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
182                             h->a53_caption[h->a53_caption_size++] = get_bits(&h->gb, 8);
183                         }
184
185                         skip_bits(&h->gb, 8);
186                     }
187                 }
188             }
189             break;
190         default:
191             skip_bits(&h->gb, size * 8);
192             break;
193     }
194
195     return 0;
196 }
197
198 static int decode_unregistered_user_data(H264Context *h, int size)
199 {
200     uint8_t user_data[16 + 256];
201     int e, build, i;
202
203     if (size < 16)
204         return AVERROR_INVALIDDATA;
205
206     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
207         user_data[i] = get_bits(&h->gb, 8);
208
209     user_data[i] = 0;
210     e = sscanf(user_data + 16, "x264 - core %d", &build);
211     if (e == 1 && build > 0)
212         h->x264_build = build;
213     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
214         h->x264_build = 67;
215
216     if (h->avctx->debug & FF_DEBUG_BUGS)
217         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
218
219     for (; i < size; i++)
220         skip_bits(&h->gb, 8);
221
222     return 0;
223 }
224
225 static int decode_recovery_point(H264Context *h)
226 {
227     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
228
229     /* 1b exact_match_flag,
230      * 1b broken_link_flag,
231      * 2b changing_slice_group_idc */
232     skip_bits(&h->gb, 4);
233
234     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
235         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
236
237     h->has_recovery_point = 1;
238
239     return 0;
240 }
241
242 static int decode_buffering_period(H264Context *h)
243 {
244     unsigned int sps_id;
245     int sched_sel_idx;
246     SPS *sps;
247
248     sps_id = get_ue_golomb_31(&h->gb);
249     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
250         av_log(h->avctx, AV_LOG_ERROR,
251                "non-existing SPS %d referenced in buffering period\n", sps_id);
252         return AVERROR_INVALIDDATA;
253     }
254     sps = h->sps_buffers[sps_id];
255
256     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
257     if (sps->nal_hrd_parameters_present_flag) {
258         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
259             h->initial_cpb_removal_delay[sched_sel_idx] =
260                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
261             // initial_cpb_removal_delay_offset
262             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
263         }
264     }
265     if (sps->vcl_hrd_parameters_present_flag) {
266         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
267             h->initial_cpb_removal_delay[sched_sel_idx] =
268                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
269             // initial_cpb_removal_delay_offset
270             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
271         }
272     }
273
274     h->sei_buffering_period_present = 1;
275     return 0;
276 }
277
278 static int decode_frame_packing_arrangement(H264Context *h)
279 {
280     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
281     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
282     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
283
284     if (h->sei_frame_packing_present) {
285         h->sei_fpa.frame_packing_arrangement_type =
286         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
287         h->sei_fpa.quincunx_sampling_flag         =
288         h->quincunx_subsampling           = get_bits1(&h->gb);
289         h->sei_fpa.content_interpretation_type    =
290         h->content_interpretation_type    = get_bits(&h->gb, 6);
291
292         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
293         // field_views_flag, current_frame_is_frame0_flag,
294         // frame0_self_contained_flag, frame1_self_contained_flag
295         skip_bits(&h->gb, 6);
296
297         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
298             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
299         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
300         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
301     }
302     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
303
304     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
305         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
306                                        h->sei_fpa.frame_packing_arrangement_id,
307                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
308                                        h->sei_fpa.frame_packing_arrangement_type,
309                                        h->sei_fpa.quincunx_sampling_flag,
310                                        h->sei_fpa.content_interpretation_type,
311                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
312
313     return 0;
314 }
315
316 static int decode_display_orientation(H264Context *h)
317 {
318     h->sei_display_orientation_present = !get_bits1(&h->gb);
319
320     if (h->sei_display_orientation_present) {
321         h->sei_hflip = get_bits1(&h->gb);     // hor_flip
322         h->sei_vflip = get_bits1(&h->gb);     // ver_flip
323
324         h->sei_anticlockwise_rotation = get_bits(&h->gb, 16);
325         get_ue_golomb(&h->gb);  // display_orientation_repetition_period
326         skip_bits1(&h->gb);     // display_orientation_extension_flag
327     }
328
329     return 0;
330 }
331
332 static int decode_GreenMetadata(H264Context *h)
333 {
334     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
335         av_log(h->avctx, AV_LOG_DEBUG,          "Green Metadata Info SEI message\n");
336
337     h->sei_green_metadata.green_metadata_type=get_bits(&h->gb, 8);
338
339     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
340         av_log(h->avctx, AV_LOG_DEBUG,          "green_metadata_type                            = %d\n",
341                h->sei_green_metadata.green_metadata_type);
342
343     if (h->sei_green_metadata.green_metadata_type==0){
344         h->sei_green_metadata.period_type=get_bits(&h->gb, 8);
345
346         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
347             av_log(h->avctx, AV_LOG_DEBUG,      "green_metadata_period_type                     = %d\n",
348                    h->sei_green_metadata.period_type);
349
350         if (h->sei_green_metadata.green_metadata_type==2){
351             h->sei_green_metadata.num_seconds = get_bits(&h->gb, 16);
352             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
353                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_seconds                     = %d\n",
354                        h->sei_green_metadata.num_seconds);
355         }
356         else if (h->sei_green_metadata.period_type==3){
357             h->sei_green_metadata.num_pictures = get_bits(&h->gb, 16);
358             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
359                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_pictures                    = %d\n",
360                        h->sei_green_metadata.num_pictures);
361         }
362
363         h->sei_green_metadata.percent_non_zero_macroblocks=get_bits(&h->gb, 8);
364         h->sei_green_metadata.percent_intra_coded_macroblocks=get_bits(&h->gb, 8);
365         h->sei_green_metadata.percent_six_tap_filtering=get_bits(&h->gb, 8);
366         h->sei_green_metadata.percent_alpha_point_deblocking_instance=get_bits(&h->gb, 8);
367
368         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
369             av_log(h->avctx, AV_LOG_DEBUG,      "SEI GREEN Complexity Metrics                   = %f %f %f %f\n",
370                                            (float)h->sei_green_metadata.percent_non_zero_macroblocks/255,
371                                            (float)h->sei_green_metadata.percent_intra_coded_macroblocks/255,
372                                            (float)h->sei_green_metadata.percent_six_tap_filtering/255,
373                                            (float)h->sei_green_metadata.percent_alpha_point_deblocking_instance/255);
374
375     }else if( h->sei_green_metadata.green_metadata_type==1){
376         h->sei_green_metadata.xsd_metric_type=get_bits(&h->gb, 8);
377         h->sei_green_metadata.xsd_metric_value=get_bits(&h->gb, 16);
378
379         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
380             av_log(h->avctx, AV_LOG_DEBUG,      "xsd_metric_type                                = %d\n",
381                    h->sei_green_metadata.xsd_metric_type);
382         if ( h->sei_green_metadata.xsd_metric_type==0){
383             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
384                 av_log(h->avctx, AV_LOG_DEBUG,  "xsd_metric_value                               = %f\n",
385                        (float)h->sei_green_metadata.xsd_metric_value/100);
386         }
387     }
388
389     return 0;
390 }
391
392 int ff_h264_decode_sei(H264Context *h)
393 {
394     while (get_bits_left(&h->gb) > 16 && show_bits(&h->gb, 16)) {
395         int type = 0;
396         unsigned size = 0;
397         unsigned next;
398         int ret  = 0;
399
400         do {
401             if (get_bits_left(&h->gb) < 8)
402                 return AVERROR_INVALIDDATA;
403             type += show_bits(&h->gb, 8);
404         } while (get_bits(&h->gb, 8) == 255);
405
406         do {
407             if (get_bits_left(&h->gb) < 8)
408                 return AVERROR_INVALIDDATA;
409             size += show_bits(&h->gb, 8);
410         } while (get_bits(&h->gb, 8) == 255);
411
412         if (h->avctx->debug&FF_DEBUG_STARTCODE)
413             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
414
415         if (size > get_bits_left(&h->gb) / 8) {
416             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
417                    type, 8*size, get_bits_left(&h->gb));
418             return AVERROR_INVALIDDATA;
419         }
420         next = get_bits_count(&h->gb) + 8 * size;
421
422         switch (type) {
423         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
424             ret = decode_picture_timing(h);
425             break;
426         case SEI_TYPE_USER_DATA_REGISTERED:
427             ret = decode_registered_user_data(h, size);
428             break;
429         case SEI_TYPE_USER_DATA_UNREGISTERED:
430             ret = decode_unregistered_user_data(h, size);
431             break;
432         case SEI_TYPE_RECOVERY_POINT:
433             ret = decode_recovery_point(h);
434             break;
435         case SEI_TYPE_BUFFERING_PERIOD:
436             ret = decode_buffering_period(h);
437             break;
438         case SEI_TYPE_FRAME_PACKING:
439             ret = decode_frame_packing_arrangement(h);
440             break;
441         case SEI_TYPE_DISPLAY_ORIENTATION:
442             ret = decode_display_orientation(h);
443             break;
444         case SEI_TYPE_GREEN_METADATA:
445             ret = decode_GreenMetadata(h);
446             break;
447         default:
448             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
449         }
450         if (ret < 0)
451             return ret;
452
453         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
454
455         // FIXME check bits here
456         align_get_bits(&h->gb);
457     }
458
459     return 0;
460 }
461
462 const char* ff_h264_sei_stereo_mode(H264Context *h)
463 {
464     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
465         switch (h->sei_fpa.frame_packing_arrangement_type) {
466             case SEI_FPA_TYPE_CHECKERBOARD:
467                 if (h->sei_fpa.content_interpretation_type == 2)
468                     return "checkerboard_rl";
469                 else
470                     return "checkerboard_lr";
471             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
472                 if (h->sei_fpa.content_interpretation_type == 2)
473                     return "col_interleaved_rl";
474                 else
475                     return "col_interleaved_lr";
476             case SEI_FPA_TYPE_INTERLEAVE_ROW:
477                 if (h->sei_fpa.content_interpretation_type == 2)
478                     return "row_interleaved_rl";
479                 else
480                     return "row_interleaved_lr";
481             case SEI_FPA_TYPE_SIDE_BY_SIDE:
482                 if (h->sei_fpa.content_interpretation_type == 2)
483                     return "right_left";
484                 else
485                     return "left_right";
486             case SEI_FPA_TYPE_TOP_BOTTOM:
487                 if (h->sei_fpa.content_interpretation_type == 2)
488                     return "bottom_top";
489                 else
490                     return "top_bottom";
491             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
492                 if (h->sei_fpa.content_interpretation_type == 2)
493                     return "block_rl";
494                 else
495                     return "block_lr";
496             case SEI_FPA_TYPE_2D:
497             default:
498                 return "mono";
499         }
500     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
501         return "mono";
502     } else {
503         return NULL;
504     }
505 }