avcodec/h264: Greenmetadata SEI parsing
[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
48 static int decode_picture_timing(H264Context *h)
49 {
50     SPS *sps = &h->sps;
51     int i;
52
53     for (i = 0; i<MAX_SPS_COUNT; i++)
54         if (!sps->log2_max_frame_num && h->sps_buffers[i])
55             sps = h->sps_buffers[i];
56
57     if (sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag) {
58         h->sei_cpb_removal_delay = get_bits_long(&h->gb,
59                                                  sps->cpb_removal_delay_length);
60         h->sei_dpb_output_delay  = get_bits_long(&h->gb,
61                                                  sps->dpb_output_delay_length);
62     }
63     if (sps->pic_struct_present_flag) {
64         unsigned int i, num_clock_ts;
65
66         h->sei_pic_struct = get_bits(&h->gb, 4);
67         h->sei_ct_type    = 0;
68
69         if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
70             return AVERROR_INVALIDDATA;
71
72         num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
73
74         for (i = 0; i < num_clock_ts; i++) {
75             if (get_bits(&h->gb, 1)) {                /* clock_timestamp_flag */
76                 unsigned int full_timestamp_flag;
77
78                 h->sei_ct_type |= 1 << get_bits(&h->gb, 2);
79                 skip_bits(&h->gb, 1);                 /* nuit_field_based_flag */
80                 skip_bits(&h->gb, 5);                 /* counting_type */
81                 full_timestamp_flag = get_bits(&h->gb, 1);
82                 skip_bits(&h->gb, 1);                 /* discontinuity_flag */
83                 skip_bits(&h->gb, 1);                 /* cnt_dropped_flag */
84                 skip_bits(&h->gb, 8);                 /* n_frames */
85                 if (full_timestamp_flag) {
86                     skip_bits(&h->gb, 6);             /* seconds_value 0..59 */
87                     skip_bits(&h->gb, 6);             /* minutes_value 0..59 */
88                     skip_bits(&h->gb, 5);             /* hours_value 0..23 */
89                 } else {
90                     if (get_bits(&h->gb, 1)) {        /* seconds_flag */
91                         skip_bits(&h->gb, 6);         /* seconds_value range 0..59 */
92                         if (get_bits(&h->gb, 1)) {    /* minutes_flag */
93                             skip_bits(&h->gb, 6);     /* minutes_value 0..59 */
94                             if (get_bits(&h->gb, 1))  /* hours_flag */
95                                 skip_bits(&h->gb, 5); /* hours_value 0..23 */
96                         }
97                     }
98                 }
99                 if (sps->time_offset_length > 0)
100                     skip_bits(&h->gb,
101                               sps->time_offset_length); /* time_offset */
102             }
103         }
104
105         if (h->avctx->debug & FF_DEBUG_PICT_INFO)
106             av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n",
107                    h->sei_ct_type, h->sei_pic_struct);
108     }
109     return 0;
110 }
111
112 static int decode_registered_user_data(H264Context *h, int size)
113 {
114     uint32_t country_code;
115     uint32_t user_identifier;
116     int dtg_active_format;
117     int flag;
118
119     if (size < 7)
120         return AVERROR_INVALIDDATA;
121     size -= 7;
122
123     country_code = get_bits(&h->gb, 8); // itu_t_t35_country_code
124     if (country_code == 0xFF) {
125         skip_bits(&h->gb, 8);           // itu_t_t35_country_code_extension_byte
126         size--;
127     }
128
129     /* itu_t_t35_payload_byte follows */
130     skip_bits(&h->gb, 8);              // terminal provider code
131     skip_bits(&h->gb, 8);              // terminal provider oriented code
132     user_identifier = get_bits_long(&h->gb, 32);
133
134     switch (user_identifier) {
135         case MKBETAG('D', 'T', 'G', '1'):       // afd_data
136             if (size-- < 1)
137                 return AVERROR_INVALIDDATA;
138             skip_bits(&h->gb, 1);               // 0
139             flag = get_bits(&h->gb, 1);         // active_format_flag
140             skip_bits(&h->gb, 6);               // reserved
141
142             if (flag) {
143                 if (size-- < 1)
144                     return AVERROR_INVALIDDATA;
145                 skip_bits(&h->gb, 4);           // reserved
146                 h->active_format_description   = get_bits(&h->gb, 4);
147                 h->sei_reguserdata_afd_present = 1;
148 #if FF_API_AFD
149 FF_DISABLE_DEPRECATION_WARNINGS
150                 h->avctx->dtg_active_format = h->active_format_description;
151 FF_ENABLE_DEPRECATION_WARNINGS
152 #endif /* FF_API_AFD */
153             }
154             break;
155         default:
156             skip_bits(&h->gb, size * 8);
157             break;
158     }
159
160     return 0;
161 }
162
163 static int decode_unregistered_user_data(H264Context *h, int size)
164 {
165     uint8_t user_data[16 + 256];
166     int e, build, i;
167
168     if (size < 16)
169         return AVERROR_INVALIDDATA;
170
171     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
172         user_data[i] = get_bits(&h->gb, 8);
173
174     user_data[i] = 0;
175     e = sscanf(user_data + 16, "x264 - core %d", &build);
176     if (e == 1 && build > 0)
177         h->x264_build = build;
178     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
179         h->x264_build = 67;
180
181     if (h->avctx->debug & FF_DEBUG_BUGS)
182         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
183
184     for (; i < size; i++)
185         skip_bits(&h->gb, 8);
186
187     return 0;
188 }
189
190 static int decode_recovery_point(H264Context *h)
191 {
192     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
193
194     /* 1b exact_match_flag,
195      * 1b broken_link_flag,
196      * 2b changing_slice_group_idc */
197     skip_bits(&h->gb, 4);
198
199     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
200         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
201
202     h->has_recovery_point = 1;
203
204     return 0;
205 }
206
207 static int decode_buffering_period(H264Context *h)
208 {
209     unsigned int sps_id;
210     int sched_sel_idx;
211     SPS *sps;
212
213     sps_id = get_ue_golomb_31(&h->gb);
214     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
215         av_log(h->avctx, AV_LOG_ERROR,
216                "non-existing SPS %d referenced in buffering period\n", sps_id);
217         return AVERROR_INVALIDDATA;
218     }
219     sps = h->sps_buffers[sps_id];
220
221     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
222     if (sps->nal_hrd_parameters_present_flag) {
223         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
224             h->initial_cpb_removal_delay[sched_sel_idx] =
225                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
226             // initial_cpb_removal_delay_offset
227             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
228         }
229     }
230     if (sps->vcl_hrd_parameters_present_flag) {
231         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
232             h->initial_cpb_removal_delay[sched_sel_idx] =
233                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
234             // initial_cpb_removal_delay_offset
235             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
236         }
237     }
238
239     h->sei_buffering_period_present = 1;
240     return 0;
241 }
242
243 static int decode_frame_packing_arrangement(H264Context *h)
244 {
245     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
246     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
247     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
248
249     if (h->sei_frame_packing_present) {
250         h->sei_fpa.frame_packing_arrangement_type =
251         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
252         h->sei_fpa.quincunx_sampling_flag         =
253         h->quincunx_subsampling           = get_bits1(&h->gb);
254         h->sei_fpa.content_interpretation_type    =
255         h->content_interpretation_type    = get_bits(&h->gb, 6);
256
257         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
258         // field_views_flag, current_frame_is_frame0_flag,
259         // frame0_self_contained_flag, frame1_self_contained_flag
260         skip_bits(&h->gb, 6);
261
262         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
263             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
264         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
265         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
266     }
267     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
268
269     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
270         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
271                                        h->sei_fpa.frame_packing_arrangement_id,
272                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
273                                        h->sei_fpa.frame_packing_arrangement_type,
274                                        h->sei_fpa.quincunx_sampling_flag,
275                                        h->sei_fpa.content_interpretation_type,
276                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
277
278     return 0;
279 }
280
281 static int decode_display_orientation(H264Context *h)
282 {
283     h->sei_display_orientation_present = !get_bits1(&h->gb);
284
285     if (h->sei_display_orientation_present) {
286         h->sei_hflip = get_bits1(&h->gb);     // hor_flip
287         h->sei_vflip = get_bits1(&h->gb);     // ver_flip
288
289         h->sei_anticlockwise_rotation = get_bits(&h->gb, 16);
290         get_ue_golomb(&h->gb);  // display_orientation_repetition_period
291         skip_bits1(&h->gb);     // display_orientation_extension_flag
292     }
293
294     return 0;
295 }
296
297 static int decode_GreenMetadata(H264Context *h)
298 {
299     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
300         av_log(h->avctx, AV_LOG_DEBUG,          "Green Metadata Info SEI message\n");
301
302     h->sei_green_metadata.green_metadata_type=get_bits(&h->gb, 8);
303
304     if (h->avctx->debug & FF_DEBUG_GREEN_MD)
305         av_log(h->avctx, AV_LOG_DEBUG,          "green_metadata_type                            = %d\n",
306                h->sei_green_metadata.green_metadata_type);
307
308     if (h->sei_green_metadata.green_metadata_type==0){
309         h->sei_green_metadata.period_type=get_bits(&h->gb, 8);
310
311         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
312             av_log(h->avctx, AV_LOG_DEBUG,      "green_metadata_period_type                     = %d\n",
313                    h->sei_green_metadata.period_type);
314
315         if (h->sei_green_metadata.green_metadata_type==2){
316             h->sei_green_metadata.num_seconds = get_bits(&h->gb, 16);
317             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
318                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_seconds                     = %d\n",
319                        h->sei_green_metadata.num_seconds);
320         }
321         else if (h->sei_green_metadata.period_type==3){
322             h->sei_green_metadata.num_pictures = get_bits(&h->gb, 16);
323             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
324                 av_log(h->avctx, AV_LOG_DEBUG,  "green_metadata_num_pictures                    = %d\n",
325                        h->sei_green_metadata.num_pictures);
326         }
327
328         h->sei_green_metadata.percent_non_zero_macroblocks=get_bits(&h->gb, 8);
329         h->sei_green_metadata.percent_intra_coded_macroblocks=get_bits(&h->gb, 8);
330         h->sei_green_metadata.percent_six_tap_filtering=get_bits(&h->gb, 8);
331         h->sei_green_metadata.percent_alpha_point_deblocking_instance=get_bits(&h->gb, 8);
332
333         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
334             av_log(h->avctx, AV_LOG_DEBUG,      "SEI GREEN Complexity Metrics                   = %f %f %f %f\n",
335                                            (float)h->sei_green_metadata.percent_non_zero_macroblocks/255,
336                                            (float)h->sei_green_metadata.percent_intra_coded_macroblocks/255,
337                                            (float)h->sei_green_metadata.percent_six_tap_filtering/255,
338                                            (float)h->sei_green_metadata.percent_alpha_point_deblocking_instance/255);
339
340     }else if( h->sei_green_metadata.green_metadata_type==1){
341         h->sei_green_metadata.xsd_metric_type=get_bits(&h->gb, 8);
342         h->sei_green_metadata.xsd_metric_value=get_bits(&h->gb, 16);
343
344         if (h->avctx->debug & FF_DEBUG_GREEN_MD)
345             av_log(h->avctx, AV_LOG_DEBUG,      "xsd_metric_type                                = %d\n",
346                    h->sei_green_metadata.xsd_metric_type);
347         if ( h->sei_green_metadata.xsd_metric_type==0){
348             if (h->avctx->debug & FF_DEBUG_GREEN_MD)
349                 av_log(h->avctx, AV_LOG_DEBUG,  "xsd_metric_value                               = %f\n",
350                        (float)h->sei_green_metadata.xsd_metric_value/100);
351         }
352     }
353
354     return 0;
355 }
356
357 int ff_h264_decode_sei(H264Context *h)
358 {
359     while (get_bits_left(&h->gb) > 16 && show_bits(&h->gb, 16)) {
360         int type = 0;
361         unsigned size = 0;
362         unsigned next;
363         int ret  = 0;
364
365         do {
366             if (get_bits_left(&h->gb) < 8)
367                 return AVERROR_INVALIDDATA;
368             type += show_bits(&h->gb, 8);
369         } while (get_bits(&h->gb, 8) == 255);
370
371         do {
372             if (get_bits_left(&h->gb) < 8)
373                 return AVERROR_INVALIDDATA;
374             size += show_bits(&h->gb, 8);
375         } while (get_bits(&h->gb, 8) == 255);
376
377         if (h->avctx->debug&FF_DEBUG_STARTCODE)
378             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
379
380         if (size > get_bits_left(&h->gb) / 8) {
381             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
382                    type, 8*size, get_bits_left(&h->gb));
383             return AVERROR_INVALIDDATA;
384         }
385         next = get_bits_count(&h->gb) + 8 * size;
386
387         switch (type) {
388         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
389             ret = decode_picture_timing(h);
390             break;
391         case SEI_TYPE_USER_DATA_REGISTERED:
392             ret = decode_registered_user_data(h, size);
393             break;
394         case SEI_TYPE_USER_DATA_UNREGISTERED:
395             ret = decode_unregistered_user_data(h, size);
396             break;
397         case SEI_TYPE_RECOVERY_POINT:
398             ret = decode_recovery_point(h);
399             break;
400         case SEI_TYPE_BUFFERING_PERIOD:
401             ret = decode_buffering_period(h);
402             break;
403         case SEI_TYPE_FRAME_PACKING:
404             ret = decode_frame_packing_arrangement(h);
405             break;
406         case SEI_TYPE_DISPLAY_ORIENTATION:
407             ret = decode_display_orientation(h);
408             break;
409         case SEI_TYPE_GREEN_METADATA:
410             ret = decode_GreenMetadata(h);
411             break;
412         default:
413             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
414         }
415         if (ret < 0)
416             return ret;
417
418         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
419
420         // FIXME check bits here
421         align_get_bits(&h->gb);
422     }
423
424     return 0;
425 }
426
427 const char* ff_h264_sei_stereo_mode(H264Context *h)
428 {
429     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
430         switch (h->sei_fpa.frame_packing_arrangement_type) {
431             case SEI_FPA_TYPE_CHECKERBOARD:
432                 if (h->sei_fpa.content_interpretation_type == 2)
433                     return "checkerboard_rl";
434                 else
435                     return "checkerboard_lr";
436             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
437                 if (h->sei_fpa.content_interpretation_type == 2)
438                     return "col_interleaved_rl";
439                 else
440                     return "col_interleaved_lr";
441             case SEI_FPA_TYPE_INTERLEAVE_ROW:
442                 if (h->sei_fpa.content_interpretation_type == 2)
443                     return "row_interleaved_rl";
444                 else
445                     return "row_interleaved_lr";
446             case SEI_FPA_TYPE_SIDE_BY_SIDE:
447                 if (h->sei_fpa.content_interpretation_type == 2)
448                     return "right_left";
449                 else
450                     return "left_right";
451             case SEI_FPA_TYPE_TOP_BOTTOM:
452                 if (h->sei_fpa.content_interpretation_type == 2)
453                     return "bottom_top";
454                 else
455                     return "top_bottom";
456             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
457                 if (h->sei_fpa.content_interpretation_type == 2)
458                     return "block_rl";
459                 else
460                     return "block_lr";
461             case SEI_FPA_TYPE_2D:
462             default:
463                 return "mono";
464         }
465     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
466         return "mono";
467     } else {
468         return NULL;
469     }
470 }