avcodec: Add support for per-frame AFD output 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 }
46
47 static int decode_picture_timing(H264Context *h)
48 {
49     SPS *sps = &h->sps;
50     int i;
51
52     for (i = 0; i<MAX_SPS_COUNT; i++)
53         if (!sps->log2_max_frame_num && h->sps_buffers[i])
54             sps = h->sps_buffers[i];
55
56     if (sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag) {
57         h->sei_cpb_removal_delay = get_bits_long(&h->gb,
58                                                  sps->cpb_removal_delay_length);
59         h->sei_dpb_output_delay  = get_bits_long(&h->gb,
60                                                  sps->dpb_output_delay_length);
61     }
62     if (sps->pic_struct_present_flag) {
63         unsigned int i, num_clock_ts;
64
65         h->sei_pic_struct = get_bits(&h->gb, 4);
66         h->sei_ct_type    = 0;
67
68         if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
69             return AVERROR_INVALIDDATA;
70
71         num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
72
73         for (i = 0; i < num_clock_ts; i++) {
74             if (get_bits(&h->gb, 1)) {                /* clock_timestamp_flag */
75                 unsigned int full_timestamp_flag;
76
77                 h->sei_ct_type |= 1 << get_bits(&h->gb, 2);
78                 skip_bits(&h->gb, 1);                 /* nuit_field_based_flag */
79                 skip_bits(&h->gb, 5);                 /* counting_type */
80                 full_timestamp_flag = get_bits(&h->gb, 1);
81                 skip_bits(&h->gb, 1);                 /* discontinuity_flag */
82                 skip_bits(&h->gb, 1);                 /* cnt_dropped_flag */
83                 skip_bits(&h->gb, 8);                 /* n_frames */
84                 if (full_timestamp_flag) {
85                     skip_bits(&h->gb, 6);             /* seconds_value 0..59 */
86                     skip_bits(&h->gb, 6);             /* minutes_value 0..59 */
87                     skip_bits(&h->gb, 5);             /* hours_value 0..23 */
88                 } else {
89                     if (get_bits(&h->gb, 1)) {        /* seconds_flag */
90                         skip_bits(&h->gb, 6);         /* seconds_value range 0..59 */
91                         if (get_bits(&h->gb, 1)) {    /* minutes_flag */
92                             skip_bits(&h->gb, 6);     /* minutes_value 0..59 */
93                             if (get_bits(&h->gb, 1))  /* hours_flag */
94                                 skip_bits(&h->gb, 5); /* hours_value 0..23 */
95                         }
96                     }
97                 }
98                 if (sps->time_offset_length > 0)
99                     skip_bits(&h->gb,
100                               sps->time_offset_length); /* time_offset */
101             }
102         }
103
104         if (h->avctx->debug & FF_DEBUG_PICT_INFO)
105             av_log(h->avctx, AV_LOG_DEBUG, "ct_type:%X pic_struct:%d\n",
106                    h->sei_ct_type, h->sei_pic_struct);
107     }
108     return 0;
109 }
110
111 static int decode_user_data_itu_t_t35(H264Context *h, int size)
112 {
113     uint32_t user_identifier;
114     int dtg_active_format;
115
116     if (size < 7)
117         return -1;
118     size -= 7;
119
120     skip_bits(&h->gb, 8);   // country_code
121     skip_bits(&h->gb, 16);  // provider_code
122     user_identifier = get_bits_long(&h->gb, 32);
123
124     switch (user_identifier) {
125         case 0x44544731:    // "DTG1" - AFD_data
126             if (size < 1)
127                 return -1;
128             skip_bits(&h->gb, 1);
129             if (get_bits(&h->gb, 1)) {
130                 skip_bits(&h->gb, 6);
131                 if (size < 2)
132                     return -1;
133                 skip_bits(&h->gb, 4);
134                 dtg_active_format = get_bits(&h->gb, 4);
135 #if FF_API_AFD
136 FF_DISABLE_DEPRECATION_WARNINGS
137                 h->avctx->dtg_active_format = dtg_active_format;
138 FF_ENABLE_DEPRECATION_WARNINGS
139 #endif /* FF_API_AFD */
140                 h->has_afd = 1;
141                 h->afd     = dtg_active_format;
142             } else {
143                 skip_bits(&h->gb, 6);
144             }
145             break;
146         default:
147             skip_bits(&h->gb, size * 8);
148             break;
149     }
150
151     return 0;
152 }
153
154 static int decode_unregistered_user_data(H264Context *h, int size)
155 {
156     uint8_t user_data[16 + 256];
157     int e, build, i;
158
159     if (size < 16)
160         return AVERROR_INVALIDDATA;
161
162     for (i = 0; i < sizeof(user_data) - 1 && i < size; i++)
163         user_data[i] = get_bits(&h->gb, 8);
164
165     user_data[i] = 0;
166     e = sscanf(user_data + 16, "x264 - core %d", &build);
167     if (e == 1 && build > 0)
168         h->x264_build = build;
169     if (e == 1 && build == 1 && !strncmp(user_data+16, "x264 - core 0000", 16))
170         h->x264_build = 67;
171
172     if (h->avctx->debug & FF_DEBUG_BUGS)
173         av_log(h->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data + 16);
174
175     for (; i < size; i++)
176         skip_bits(&h->gb, 8);
177
178     return 0;
179 }
180
181 static int decode_recovery_point(H264Context *h)
182 {
183     h->sei_recovery_frame_cnt = get_ue_golomb(&h->gb);
184
185     /* 1b exact_match_flag,
186      * 1b broken_link_flag,
187      * 2b changing_slice_group_idc */
188     skip_bits(&h->gb, 4);
189
190     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
191         av_log(h->avctx, AV_LOG_DEBUG, "sei_recovery_frame_cnt: %d\n", h->sei_recovery_frame_cnt);
192
193     h->has_recovery_point = 1;
194
195     return 0;
196 }
197
198 static int decode_buffering_period(H264Context *h)
199 {
200     unsigned int sps_id;
201     int sched_sel_idx;
202     SPS *sps;
203
204     sps_id = get_ue_golomb_31(&h->gb);
205     if (sps_id > 31 || !h->sps_buffers[sps_id]) {
206         av_log(h->avctx, AV_LOG_ERROR,
207                "non-existing SPS %d referenced in buffering period\n", sps_id);
208         return AVERROR_INVALIDDATA;
209     }
210     sps = h->sps_buffers[sps_id];
211
212     // NOTE: This is really so duplicated in the standard... See H.264, D.1.1
213     if (sps->nal_hrd_parameters_present_flag) {
214         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
215             h->initial_cpb_removal_delay[sched_sel_idx] =
216                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
217             // initial_cpb_removal_delay_offset
218             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
219         }
220     }
221     if (sps->vcl_hrd_parameters_present_flag) {
222         for (sched_sel_idx = 0; sched_sel_idx < sps->cpb_cnt; sched_sel_idx++) {
223             h->initial_cpb_removal_delay[sched_sel_idx] =
224                 get_bits_long(&h->gb, sps->initial_cpb_removal_delay_length);
225             // initial_cpb_removal_delay_offset
226             skip_bits(&h->gb, sps->initial_cpb_removal_delay_length);
227         }
228     }
229
230     h->sei_buffering_period_present = 1;
231     return 0;
232 }
233
234 static int decode_frame_packing_arrangement(H264Context *h)
235 {
236     h->sei_fpa.frame_packing_arrangement_id          = get_ue_golomb(&h->gb);
237     h->sei_fpa.frame_packing_arrangement_cancel_flag = get_bits1(&h->gb);
238     h->sei_frame_packing_present = !h->sei_fpa.frame_packing_arrangement_cancel_flag;
239
240     if (h->sei_frame_packing_present) {
241         h->sei_fpa.frame_packing_arrangement_type =
242         h->frame_packing_arrangement_type = get_bits(&h->gb, 7);
243         h->sei_fpa.quincunx_sampling_flag         =
244         h->quincunx_subsampling           = get_bits1(&h->gb);
245         h->sei_fpa.content_interpretation_type    =
246         h->content_interpretation_type    = get_bits(&h->gb, 6);
247
248         // the following skips: spatial_flipping_flag, frame0_flipped_flag,
249         // field_views_flag, current_frame_is_frame0_flag,
250         // frame0_self_contained_flag, frame1_self_contained_flag
251         skip_bits(&h->gb, 6);
252
253         if (!h->quincunx_subsampling && h->frame_packing_arrangement_type != 5)
254             skip_bits(&h->gb, 16);      // frame[01]_grid_position_[xy]
255         skip_bits(&h->gb, 8);           // frame_packing_arrangement_reserved_byte
256         h->sei_fpa.frame_packing_arrangement_repetition_period = get_ue_golomb(&h->gb) /* frame_packing_arrangement_repetition_period */;
257     }
258     skip_bits1(&h->gb);                 // frame_packing_arrangement_extension_flag
259
260     if (h->avctx->debug & FF_DEBUG_PICT_INFO)
261         av_log(h->avctx, AV_LOG_DEBUG, "SEI FPA %d %d %d %d %d %d\n",
262                                        h->sei_fpa.frame_packing_arrangement_id,
263                                        h->sei_fpa.frame_packing_arrangement_cancel_flag,
264                                        h->sei_fpa.frame_packing_arrangement_type,
265                                        h->sei_fpa.quincunx_sampling_flag,
266                                        h->sei_fpa.content_interpretation_type,
267                                        h->sei_fpa.frame_packing_arrangement_repetition_period);
268
269     return 0;
270 }
271
272 static int decode_display_orientation(H264Context *h)
273 {
274     h->sei_display_orientation_present = !get_bits1(&h->gb);
275
276     if (h->sei_display_orientation_present) {
277         h->sei_hflip = get_bits1(&h->gb);     // hor_flip
278         h->sei_vflip = get_bits1(&h->gb);     // ver_flip
279
280         h->sei_anticlockwise_rotation = get_bits(&h->gb, 16);
281         get_ue_golomb(&h->gb);  // display_orientation_repetition_period
282         skip_bits1(&h->gb);     // display_orientation_extension_flag
283     }
284
285     return 0;
286 }
287
288 int ff_h264_decode_sei(H264Context *h)
289 {
290     while (get_bits_left(&h->gb) > 16 && show_bits(&h->gb, 16)) {
291         int type = 0;
292         unsigned size = 0;
293         unsigned next;
294         int ret  = 0;
295
296         do {
297             if (get_bits_left(&h->gb) < 8)
298                 return AVERROR_INVALIDDATA;
299             type += show_bits(&h->gb, 8);
300         } while (get_bits(&h->gb, 8) == 255);
301
302         do {
303             if (get_bits_left(&h->gb) < 8)
304                 return AVERROR_INVALIDDATA;
305             size += show_bits(&h->gb, 8);
306         } while (get_bits(&h->gb, 8) == 255);
307
308         if (h->avctx->debug&FF_DEBUG_STARTCODE)
309             av_log(h->avctx, AV_LOG_DEBUG, "SEI %d len:%d\n", type, size);
310
311         if (size > get_bits_left(&h->gb) / 8) {
312             av_log(h->avctx, AV_LOG_ERROR, "SEI type %d size %d truncated at %d\n",
313                    type, 8*size, get_bits_left(&h->gb));
314             return AVERROR_INVALIDDATA;
315         }
316         next = get_bits_count(&h->gb) + 8 * size;
317
318         switch (type) {
319         case SEI_TYPE_PIC_TIMING: // Picture timing SEI
320             ret = decode_picture_timing(h);
321             if (ret < 0)
322                 return ret;
323             break;
324         case SEI_TYPE_USER_DATA_ITU_T_T35:
325             if (decode_user_data_itu_t_t35(h, size) < 0)
326                 return -1;
327             break;
328         case SEI_TYPE_USER_DATA_UNREGISTERED:
329             ret = decode_unregistered_user_data(h, size);
330             if (ret < 0)
331                 return ret;
332             break;
333         case SEI_TYPE_RECOVERY_POINT:
334             ret = decode_recovery_point(h);
335             if (ret < 0)
336                 return ret;
337             break;
338         case SEI_TYPE_BUFFERING_PERIOD:
339             ret = decode_buffering_period(h);
340             if (ret < 0)
341                 return ret;
342             break;
343         case SEI_TYPE_FRAME_PACKING:
344             ret = decode_frame_packing_arrangement(h);
345             if (ret < 0)
346                 return ret;
347             break;
348         case SEI_TYPE_DISPLAY_ORIENTATION:
349             ret = decode_display_orientation(h);
350             if (ret < 0)
351                 return ret;
352             break;
353         default:
354             av_log(h->avctx, AV_LOG_DEBUG, "unknown SEI type %d\n", type);
355         }
356         skip_bits_long(&h->gb, next - get_bits_count(&h->gb));
357
358         // FIXME check bits here
359         align_get_bits(&h->gb);
360     }
361
362     return 0;
363 }
364
365 const char* ff_h264_sei_stereo_mode(H264Context *h)
366 {
367     if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 0) {
368         switch (h->sei_fpa.frame_packing_arrangement_type) {
369             case SEI_FPA_TYPE_CHECKERBOARD:
370                 if (h->sei_fpa.content_interpretation_type == 2)
371                     return "checkerboard_rl";
372                 else
373                     return "checkerboard_lr";
374             case SEI_FPA_TYPE_INTERLEAVE_COLUMN:
375                 if (h->sei_fpa.content_interpretation_type == 2)
376                     return "col_interleaved_rl";
377                 else
378                     return "col_interleaved_lr";
379             case SEI_FPA_TYPE_INTERLEAVE_ROW:
380                 if (h->sei_fpa.content_interpretation_type == 2)
381                     return "row_interleaved_rl";
382                 else
383                     return "row_interleaved_lr";
384             case SEI_FPA_TYPE_SIDE_BY_SIDE:
385                 if (h->sei_fpa.content_interpretation_type == 2)
386                     return "right_left";
387                 else
388                     return "left_right";
389             case SEI_FPA_TYPE_TOP_BOTTOM:
390                 if (h->sei_fpa.content_interpretation_type == 2)
391                     return "bottom_top";
392                 else
393                     return "top_bottom";
394             case SEI_FPA_TYPE_INTERLEAVE_TEMPORAL:
395                 if (h->sei_fpa.content_interpretation_type == 2)
396                     return "block_rl";
397                 else
398                     return "block_lr";
399             case SEI_FPA_TYPE_2D:
400             default:
401                 return "mono";
402         }
403     } else if (h->sei_fpa.frame_packing_arrangement_cancel_flag == 1) {
404         return "mono";
405     } else {
406         return NULL;
407     }
408 }