summaryrefslogtreecommitdiff
path: root/libavformat/hevc.c (plain)
blob: 1a2d6cdd2b2587bfac2d8ef7ae53880910408459
1/*
2 * Copyright (c) 2014 Tim Walker <tdskywalker@gmail.com>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21#include "libavcodec/avcodec.h"
22#include "libavcodec/get_bits.h"
23#include "libavcodec/golomb.h"
24#include "libavcodec/hevc.h"
25#include "libavutil/intreadwrite.h"
26#include "avc.h"
27#include "avio.h"
28#include "hevc.h"
29
30#define MAX_SPATIAL_SEGMENTATION 4096 // max. value of u(12) field
31
32typedef struct HVCCNALUnitArray {
33 uint8_t array_completeness;
34 uint8_t NAL_unit_type;
35 uint16_t numNalus;
36 uint16_t *nalUnitLength;
37 uint8_t **nalUnit;
38} HVCCNALUnitArray;
39
40typedef struct HEVCDecoderConfigurationRecord {
41 uint8_t configurationVersion;
42 uint8_t general_profile_space;
43 uint8_t general_tier_flag;
44 uint8_t general_profile_idc;
45 uint32_t general_profile_compatibility_flags;
46 uint64_t general_constraint_indicator_flags;
47 uint8_t general_level_idc;
48 uint16_t min_spatial_segmentation_idc;
49 uint8_t parallelismType;
50 uint8_t chromaFormat;
51 uint8_t bitDepthLumaMinus8;
52 uint8_t bitDepthChromaMinus8;
53 uint16_t avgFrameRate;
54 uint8_t constantFrameRate;
55 uint8_t numTemporalLayers;
56 uint8_t temporalIdNested;
57 uint8_t lengthSizeMinusOne;
58 uint8_t numOfArrays;
59 HVCCNALUnitArray *array;
60} HEVCDecoderConfigurationRecord;
61
62typedef struct HVCCProfileTierLevel {
63 uint8_t profile_space;
64 uint8_t tier_flag;
65 uint8_t profile_idc;
66 uint32_t profile_compatibility_flags;
67 uint64_t constraint_indicator_flags;
68 uint8_t level_idc;
69} HVCCProfileTierLevel;
70
71static void hvcc_update_ptl(HEVCDecoderConfigurationRecord *hvcc,
72 HVCCProfileTierLevel *ptl)
73{
74 /*
75 * The value of general_profile_space in all the parameter sets must be
76 * identical.
77 */
78 hvcc->general_profile_space = ptl->profile_space;
79
80 /*
81 * The level indication general_level_idc must indicate a level of
82 * capability equal to or greater than the highest level indicated for the
83 * highest tier in all the parameter sets.
84 */
85 if (hvcc->general_tier_flag < ptl->tier_flag)
86 hvcc->general_level_idc = ptl->level_idc;
87 else
88 hvcc->general_level_idc = FFMAX(hvcc->general_level_idc, ptl->level_idc);
89
90 /*
91 * The tier indication general_tier_flag must indicate a tier equal to or
92 * greater than the highest tier indicated in all the parameter sets.
93 */
94 hvcc->general_tier_flag = FFMAX(hvcc->general_tier_flag, ptl->tier_flag);
95
96 /*
97 * The profile indication general_profile_idc must indicate a profile to
98 * which the stream associated with this configuration record conforms.
99 *
100 * If the sequence parameter sets are marked with different profiles, then
101 * the stream may need examination to determine which profile, if any, the
102 * entire stream conforms to. If the entire stream is not examined, or the
103 * examination reveals that there is no profile to which the entire stream
104 * conforms, then the entire stream must be split into two or more
105 * sub-streams with separate configuration records in which these rules can
106 * be met.
107 *
108 * Note: set the profile to the highest value for the sake of simplicity.
109 */
110 hvcc->general_profile_idc = FFMAX(hvcc->general_profile_idc, ptl->profile_idc);
111
112 /*
113 * Each bit in general_profile_compatibility_flags may only be set if all
114 * the parameter sets set that bit.
115 */
116 hvcc->general_profile_compatibility_flags &= ptl->profile_compatibility_flags;
117
118 /*
119 * Each bit in general_constraint_indicator_flags may only be set if all
120 * the parameter sets set that bit.
121 */
122 hvcc->general_constraint_indicator_flags &= ptl->constraint_indicator_flags;
123}
124
125static void hvcc_parse_ptl(GetBitContext *gb,
126 HEVCDecoderConfigurationRecord *hvcc,
127 unsigned int max_sub_layers_minus1)
128{
129 unsigned int i;
130 HVCCProfileTierLevel general_ptl;
131 uint8_t sub_layer_profile_present_flag[HEVC_MAX_SUB_LAYERS];
132 uint8_t sub_layer_level_present_flag[HEVC_MAX_SUB_LAYERS];
133
134 general_ptl.profile_space = get_bits(gb, 2);
135 general_ptl.tier_flag = get_bits1(gb);
136 general_ptl.profile_idc = get_bits(gb, 5);
137 general_ptl.profile_compatibility_flags = get_bits_long(gb, 32);
138 general_ptl.constraint_indicator_flags = get_bits64(gb, 48);
139 general_ptl.level_idc = get_bits(gb, 8);
140 hvcc_update_ptl(hvcc, &general_ptl);
141
142 for (i = 0; i < max_sub_layers_minus1; i++) {
143 sub_layer_profile_present_flag[i] = get_bits1(gb);
144 sub_layer_level_present_flag[i] = get_bits1(gb);
145 }
146
147 if (max_sub_layers_minus1 > 0)
148 for (i = max_sub_layers_minus1; i < 8; i++)
149 skip_bits(gb, 2); // reserved_zero_2bits[i]
150
151 for (i = 0; i < max_sub_layers_minus1; i++) {
152 if (sub_layer_profile_present_flag[i]) {
153 /*
154 * sub_layer_profile_space[i] u(2)
155 * sub_layer_tier_flag[i] u(1)
156 * sub_layer_profile_idc[i] u(5)
157 * sub_layer_profile_compatibility_flag[i][0..31] u(32)
158 * sub_layer_progressive_source_flag[i] u(1)
159 * sub_layer_interlaced_source_flag[i] u(1)
160 * sub_layer_non_packed_constraint_flag[i] u(1)
161 * sub_layer_frame_only_constraint_flag[i] u(1)
162 * sub_layer_reserved_zero_44bits[i] u(44)
163 */
164 skip_bits_long(gb, 32);
165 skip_bits_long(gb, 32);
166 skip_bits (gb, 24);
167 }
168
169 if (sub_layer_level_present_flag[i])
170 skip_bits(gb, 8);
171 }
172}
173
174static void skip_sub_layer_hrd_parameters(GetBitContext *gb,
175 unsigned int cpb_cnt_minus1,
176 uint8_t sub_pic_hrd_params_present_flag)
177{
178 unsigned int i;
179
180 for (i = 0; i <= cpb_cnt_minus1; i++) {
181 get_ue_golomb_long(gb); // bit_rate_value_minus1
182 get_ue_golomb_long(gb); // cpb_size_value_minus1
183
184 if (sub_pic_hrd_params_present_flag) {
185 get_ue_golomb_long(gb); // cpb_size_du_value_minus1
186 get_ue_golomb_long(gb); // bit_rate_du_value_minus1
187 }
188
189 skip_bits1(gb); // cbr_flag
190 }
191}
192
193static int skip_hrd_parameters(GetBitContext *gb, uint8_t cprms_present_flag,
194 unsigned int max_sub_layers_minus1)
195{
196 unsigned int i;
197 uint8_t sub_pic_hrd_params_present_flag = 0;
198 uint8_t nal_hrd_parameters_present_flag = 0;
199 uint8_t vcl_hrd_parameters_present_flag = 0;
200
201 if (cprms_present_flag) {
202 nal_hrd_parameters_present_flag = get_bits1(gb);
203 vcl_hrd_parameters_present_flag = get_bits1(gb);
204
205 if (nal_hrd_parameters_present_flag ||
206 vcl_hrd_parameters_present_flag) {
207 sub_pic_hrd_params_present_flag = get_bits1(gb);
208
209 if (sub_pic_hrd_params_present_flag)
210 /*
211 * tick_divisor_minus2 u(8)
212 * du_cpb_removal_delay_increment_length_minus1 u(5)
213 * sub_pic_cpb_params_in_pic_timing_sei_flag u(1)
214 * dpb_output_delay_du_length_minus1 u(5)
215 */
216 skip_bits(gb, 19);
217
218 /*
219 * bit_rate_scale u(4)
220 * cpb_size_scale u(4)
221 */
222 skip_bits(gb, 8);
223
224 if (sub_pic_hrd_params_present_flag)
225 skip_bits(gb, 4); // cpb_size_du_scale
226
227 /*
228 * initial_cpb_removal_delay_length_minus1 u(5)
229 * au_cpb_removal_delay_length_minus1 u(5)
230 * dpb_output_delay_length_minus1 u(5)
231 */
232 skip_bits(gb, 15);
233 }
234 }
235
236 for (i = 0; i <= max_sub_layers_minus1; i++) {
237 unsigned int cpb_cnt_minus1 = 0;
238 uint8_t low_delay_hrd_flag = 0;
239 uint8_t fixed_pic_rate_within_cvs_flag = 0;
240 uint8_t fixed_pic_rate_general_flag = get_bits1(gb);
241
242 if (!fixed_pic_rate_general_flag)
243 fixed_pic_rate_within_cvs_flag = get_bits1(gb);
244
245 if (fixed_pic_rate_within_cvs_flag)
246 get_ue_golomb_long(gb); // elemental_duration_in_tc_minus1
247 else
248 low_delay_hrd_flag = get_bits1(gb);
249
250 if (!low_delay_hrd_flag) {
251 cpb_cnt_minus1 = get_ue_golomb_long(gb);
252 if (cpb_cnt_minus1 > 31)
253 return AVERROR_INVALIDDATA;
254 }
255
256 if (nal_hrd_parameters_present_flag)
257 skip_sub_layer_hrd_parameters(gb, cpb_cnt_minus1,
258 sub_pic_hrd_params_present_flag);
259
260 if (vcl_hrd_parameters_present_flag)
261 skip_sub_layer_hrd_parameters(gb, cpb_cnt_minus1,
262 sub_pic_hrd_params_present_flag);
263 }
264
265 return 0;
266}
267
268static void skip_timing_info(GetBitContext *gb)
269{
270 skip_bits_long(gb, 32); // num_units_in_tick
271 skip_bits_long(gb, 32); // time_scale
272
273 if (get_bits1(gb)) // poc_proportional_to_timing_flag
274 get_ue_golomb_long(gb); // num_ticks_poc_diff_one_minus1
275}
276
277static void hvcc_parse_vui(GetBitContext *gb,
278 HEVCDecoderConfigurationRecord *hvcc,
279 unsigned int max_sub_layers_minus1)
280{
281 unsigned int min_spatial_segmentation_idc;
282
283 if (get_bits1(gb)) // aspect_ratio_info_present_flag
284 if (get_bits(gb, 8) == 255) // aspect_ratio_idc
285 skip_bits_long(gb, 32); // sar_width u(16), sar_height u(16)
286
287 if (get_bits1(gb)) // overscan_info_present_flag
288 skip_bits1(gb); // overscan_appropriate_flag
289
290 if (get_bits1(gb)) { // video_signal_type_present_flag
291 skip_bits(gb, 4); // video_format u(3), video_full_range_flag u(1)
292
293 if (get_bits1(gb)) // colour_description_present_flag
294 /*
295 * colour_primaries u(8)
296 * transfer_characteristics u(8)
297 * matrix_coeffs u(8)
298 */
299 skip_bits(gb, 24);
300 }
301
302 if (get_bits1(gb)) { // chroma_loc_info_present_flag
303 get_ue_golomb_long(gb); // chroma_sample_loc_type_top_field
304 get_ue_golomb_long(gb); // chroma_sample_loc_type_bottom_field
305 }
306
307 /*
308 * neutral_chroma_indication_flag u(1)
309 * field_seq_flag u(1)
310 * frame_field_info_present_flag u(1)
311 */
312 skip_bits(gb, 3);
313
314 if (get_bits1(gb)) { // default_display_window_flag
315 get_ue_golomb_long(gb); // def_disp_win_left_offset
316 get_ue_golomb_long(gb); // def_disp_win_right_offset
317 get_ue_golomb_long(gb); // def_disp_win_top_offset
318 get_ue_golomb_long(gb); // def_disp_win_bottom_offset
319 }
320
321 if (get_bits1(gb)) { // vui_timing_info_present_flag
322 skip_timing_info(gb);
323
324 if (get_bits1(gb)) // vui_hrd_parameters_present_flag
325 skip_hrd_parameters(gb, 1, max_sub_layers_minus1);
326 }
327
328 if (get_bits1(gb)) { // bitstream_restriction_flag
329 /*
330 * tiles_fixed_structure_flag u(1)
331 * motion_vectors_over_pic_boundaries_flag u(1)
332 * restricted_ref_pic_lists_flag u(1)
333 */
334 skip_bits(gb, 3);
335
336 min_spatial_segmentation_idc = get_ue_golomb_long(gb);
337
338 /*
339 * unsigned int(12) min_spatial_segmentation_idc;
340 *
341 * The min_spatial_segmentation_idc indication must indicate a level of
342 * spatial segmentation equal to or less than the lowest level of
343 * spatial segmentation indicated in all the parameter sets.
344 */
345 hvcc->min_spatial_segmentation_idc = FFMIN(hvcc->min_spatial_segmentation_idc,
346 min_spatial_segmentation_idc);
347
348 get_ue_golomb_long(gb); // max_bytes_per_pic_denom
349 get_ue_golomb_long(gb); // max_bits_per_min_cu_denom
350 get_ue_golomb_long(gb); // log2_max_mv_length_horizontal
351 get_ue_golomb_long(gb); // log2_max_mv_length_vertical
352 }
353}
354
355static void skip_sub_layer_ordering_info(GetBitContext *gb)
356{
357 get_ue_golomb_long(gb); // max_dec_pic_buffering_minus1
358 get_ue_golomb_long(gb); // max_num_reorder_pics
359 get_ue_golomb_long(gb); // max_latency_increase_plus1
360}
361
362static int hvcc_parse_vps(GetBitContext *gb,
363 HEVCDecoderConfigurationRecord *hvcc)
364{
365 unsigned int vps_max_sub_layers_minus1;
366
367 /*
368 * vps_video_parameter_set_id u(4)
369 * vps_reserved_three_2bits u(2)
370 * vps_max_layers_minus1 u(6)
371 */
372 skip_bits(gb, 12);
373
374 vps_max_sub_layers_minus1 = get_bits(gb, 3);
375
376 /*
377 * numTemporalLayers greater than 1 indicates that the stream to which this
378 * configuration record applies is temporally scalable and the contained
379 * number of temporal layers (also referred to as temporal sub-layer or
380 * sub-layer in ISO/IEC 23008-2) is equal to numTemporalLayers. Value 1
381 * indicates that the stream is not temporally scalable. Value 0 indicates
382 * that it is unknown whether the stream is temporally scalable.
383 */
384 hvcc->numTemporalLayers = FFMAX(hvcc->numTemporalLayers,
385 vps_max_sub_layers_minus1 + 1);
386
387 /*
388 * vps_temporal_id_nesting_flag u(1)
389 * vps_reserved_0xffff_16bits u(16)
390 */
391 skip_bits(gb, 17);
392
393 hvcc_parse_ptl(gb, hvcc, vps_max_sub_layers_minus1);
394
395 /* nothing useful for hvcC past this point */
396 return 0;
397}
398
399static void skip_scaling_list_data(GetBitContext *gb)
400{
401 int i, j, k, num_coeffs;
402
403 for (i = 0; i < 4; i++)
404 for (j = 0; j < (i == 3 ? 2 : 6); j++)
405 if (!get_bits1(gb)) // scaling_list_pred_mode_flag[i][j]
406 get_ue_golomb_long(gb); // scaling_list_pred_matrix_id_delta[i][j]
407 else {
408 num_coeffs = FFMIN(64, 1 << (4 + (i << 1)));
409
410 if (i > 1)
411 get_se_golomb_long(gb); // scaling_list_dc_coef_minus8[i-2][j]
412
413 for (k = 0; k < num_coeffs; k++)
414 get_se_golomb_long(gb); // scaling_list_delta_coef
415 }
416}
417
418static int parse_rps(GetBitContext *gb, unsigned int rps_idx,
419 unsigned int num_rps,
420 unsigned int num_delta_pocs[HEVC_MAX_SHORT_TERM_RPS_COUNT])
421{
422 unsigned int i;
423
424 if (rps_idx && get_bits1(gb)) { // inter_ref_pic_set_prediction_flag
425 /* this should only happen for slice headers, and this isn't one */
426 if (rps_idx >= num_rps)
427 return AVERROR_INVALIDDATA;
428
429 skip_bits1 (gb); // delta_rps_sign
430 get_ue_golomb_long(gb); // abs_delta_rps_minus1
431
432 num_delta_pocs[rps_idx] = 0;
433
434 /*
435 * From libavcodec/hevc_ps.c:
436 *
437 * if (is_slice_header) {
438 * //foo
439 * } else
440 * rps_ridx = &sps->st_rps[rps - sps->st_rps - 1];
441 *
442 * where:
443 * rps: &sps->st_rps[rps_idx]
444 * sps->st_rps: &sps->st_rps[0]
445 * is_slice_header: rps_idx == num_rps
446 *
447 * thus:
448 * if (num_rps != rps_idx)
449 * rps_ridx = &sps->st_rps[rps_idx - 1];
450 *
451 * NumDeltaPocs[RefRpsIdx]: num_delta_pocs[rps_idx - 1]
452 */
453 for (i = 0; i <= num_delta_pocs[rps_idx - 1]; i++) {
454 uint8_t use_delta_flag = 0;
455 uint8_t used_by_curr_pic_flag = get_bits1(gb);
456 if (!used_by_curr_pic_flag)
457 use_delta_flag = get_bits1(gb);
458
459 if (used_by_curr_pic_flag || use_delta_flag)
460 num_delta_pocs[rps_idx]++;
461 }
462 } else {
463 unsigned int num_negative_pics = get_ue_golomb_long(gb);
464 unsigned int num_positive_pics = get_ue_golomb_long(gb);
465
466 if ((num_positive_pics + (uint64_t)num_negative_pics) * 2 > get_bits_left(gb))
467 return AVERROR_INVALIDDATA;
468
469 num_delta_pocs[rps_idx] = num_negative_pics + num_positive_pics;
470
471 for (i = 0; i < num_negative_pics; i++) {
472 get_ue_golomb_long(gb); // delta_poc_s0_minus1[rps_idx]
473 skip_bits1 (gb); // used_by_curr_pic_s0_flag[rps_idx]
474 }
475
476 for (i = 0; i < num_positive_pics; i++) {
477 get_ue_golomb_long(gb); // delta_poc_s1_minus1[rps_idx]
478 skip_bits1 (gb); // used_by_curr_pic_s1_flag[rps_idx]
479 }
480 }
481
482 return 0;
483}
484
485static int hvcc_parse_sps(GetBitContext *gb,
486 HEVCDecoderConfigurationRecord *hvcc)
487{
488 unsigned int i, sps_max_sub_layers_minus1, log2_max_pic_order_cnt_lsb_minus4;
489 unsigned int num_short_term_ref_pic_sets, num_delta_pocs[HEVC_MAX_SHORT_TERM_RPS_COUNT];
490
491 skip_bits(gb, 4); // sps_video_parameter_set_id
492
493 sps_max_sub_layers_minus1 = get_bits (gb, 3);
494
495 /*
496 * numTemporalLayers greater than 1 indicates that the stream to which this
497 * configuration record applies is temporally scalable and the contained
498 * number of temporal layers (also referred to as temporal sub-layer or
499 * sub-layer in ISO/IEC 23008-2) is equal to numTemporalLayers. Value 1
500 * indicates that the stream is not temporally scalable. Value 0 indicates
501 * that it is unknown whether the stream is temporally scalable.
502 */
503 hvcc->numTemporalLayers = FFMAX(hvcc->numTemporalLayers,
504 sps_max_sub_layers_minus1 + 1);
505
506 hvcc->temporalIdNested = get_bits1(gb);
507
508 hvcc_parse_ptl(gb, hvcc, sps_max_sub_layers_minus1);
509
510 get_ue_golomb_long(gb); // sps_seq_parameter_set_id
511
512 hvcc->chromaFormat = get_ue_golomb_long(gb);
513
514 if (hvcc->chromaFormat == 3)
515 skip_bits1(gb); // separate_colour_plane_flag
516
517 get_ue_golomb_long(gb); // pic_width_in_luma_samples
518 get_ue_golomb_long(gb); // pic_height_in_luma_samples
519
520 if (get_bits1(gb)) { // conformance_window_flag
521 get_ue_golomb_long(gb); // conf_win_left_offset
522 get_ue_golomb_long(gb); // conf_win_right_offset
523 get_ue_golomb_long(gb); // conf_win_top_offset
524 get_ue_golomb_long(gb); // conf_win_bottom_offset
525 }
526
527 hvcc->bitDepthLumaMinus8 = get_ue_golomb_long(gb);
528 hvcc->bitDepthChromaMinus8 = get_ue_golomb_long(gb);
529 log2_max_pic_order_cnt_lsb_minus4 = get_ue_golomb_long(gb);
530
531 /* sps_sub_layer_ordering_info_present_flag */
532 i = get_bits1(gb) ? 0 : sps_max_sub_layers_minus1;
533 for (; i <= sps_max_sub_layers_minus1; i++)
534 skip_sub_layer_ordering_info(gb);
535
536 get_ue_golomb_long(gb); // log2_min_luma_coding_block_size_minus3
537 get_ue_golomb_long(gb); // log2_diff_max_min_luma_coding_block_size
538 get_ue_golomb_long(gb); // log2_min_transform_block_size_minus2
539 get_ue_golomb_long(gb); // log2_diff_max_min_transform_block_size
540 get_ue_golomb_long(gb); // max_transform_hierarchy_depth_inter
541 get_ue_golomb_long(gb); // max_transform_hierarchy_depth_intra
542
543 if (get_bits1(gb) && // scaling_list_enabled_flag
544 get_bits1(gb)) // sps_scaling_list_data_present_flag
545 skip_scaling_list_data(gb);
546
547 skip_bits1(gb); // amp_enabled_flag
548 skip_bits1(gb); // sample_adaptive_offset_enabled_flag
549
550 if (get_bits1(gb)) { // pcm_enabled_flag
551 skip_bits (gb, 4); // pcm_sample_bit_depth_luma_minus1
552 skip_bits (gb, 4); // pcm_sample_bit_depth_chroma_minus1
553 get_ue_golomb_long(gb); // log2_min_pcm_luma_coding_block_size_minus3
554 get_ue_golomb_long(gb); // log2_diff_max_min_pcm_luma_coding_block_size
555 skip_bits1 (gb); // pcm_loop_filter_disabled_flag
556 }
557
558 num_short_term_ref_pic_sets = get_ue_golomb_long(gb);
559 if (num_short_term_ref_pic_sets > HEVC_MAX_SHORT_TERM_RPS_COUNT)
560 return AVERROR_INVALIDDATA;
561
562 for (i = 0; i < num_short_term_ref_pic_sets; i++) {
563 int ret = parse_rps(gb, i, num_short_term_ref_pic_sets, num_delta_pocs);
564 if (ret < 0)
565 return ret;
566 }
567
568 if (get_bits1(gb)) { // long_term_ref_pics_present_flag
569 unsigned num_long_term_ref_pics_sps = get_ue_golomb_long(gb);
570 if (num_long_term_ref_pics_sps > 31U)
571 return AVERROR_INVALIDDATA;
572 for (i = 0; i < num_long_term_ref_pics_sps; i++) { // num_long_term_ref_pics_sps
573 int len = FFMIN(log2_max_pic_order_cnt_lsb_minus4 + 4, 16);
574 skip_bits (gb, len); // lt_ref_pic_poc_lsb_sps[i]
575 skip_bits1(gb); // used_by_curr_pic_lt_sps_flag[i]
576 }
577 }
578
579 skip_bits1(gb); // sps_temporal_mvp_enabled_flag
580 skip_bits1(gb); // strong_intra_smoothing_enabled_flag
581
582 if (get_bits1(gb)) // vui_parameters_present_flag
583 hvcc_parse_vui(gb, hvcc, sps_max_sub_layers_minus1);
584
585 /* nothing useful for hvcC past this point */
586 return 0;
587}
588
589static int hvcc_parse_pps(GetBitContext *gb,
590 HEVCDecoderConfigurationRecord *hvcc)
591{
592 uint8_t tiles_enabled_flag, entropy_coding_sync_enabled_flag;
593
594 get_ue_golomb_long(gb); // pps_pic_parameter_set_id
595 get_ue_golomb_long(gb); // pps_seq_parameter_set_id
596
597 /*
598 * dependent_slice_segments_enabled_flag u(1)
599 * output_flag_present_flag u(1)
600 * num_extra_slice_header_bits u(3)
601 * sign_data_hiding_enabled_flag u(1)
602 * cabac_init_present_flag u(1)
603 */
604 skip_bits(gb, 7);
605
606 get_ue_golomb_long(gb); // num_ref_idx_l0_default_active_minus1
607 get_ue_golomb_long(gb); // num_ref_idx_l1_default_active_minus1
608 get_se_golomb_long(gb); // init_qp_minus26
609
610 /*
611 * constrained_intra_pred_flag u(1)
612 * transform_skip_enabled_flag u(1)
613 */
614 skip_bits(gb, 2);
615
616 if (get_bits1(gb)) // cu_qp_delta_enabled_flag
617 get_ue_golomb_long(gb); // diff_cu_qp_delta_depth
618
619 get_se_golomb_long(gb); // pps_cb_qp_offset
620 get_se_golomb_long(gb); // pps_cr_qp_offset
621
622 /*
623 * pps_slice_chroma_qp_offsets_present_flag u(1)
624 * weighted_pred_flag u(1)
625 * weighted_bipred_flag u(1)
626 * transquant_bypass_enabled_flag u(1)
627 */
628 skip_bits(gb, 4);
629
630 tiles_enabled_flag = get_bits1(gb);
631 entropy_coding_sync_enabled_flag = get_bits1(gb);
632
633 if (entropy_coding_sync_enabled_flag && tiles_enabled_flag)
634 hvcc->parallelismType = 0; // mixed-type parallel decoding
635 else if (entropy_coding_sync_enabled_flag)
636 hvcc->parallelismType = 3; // wavefront-based parallel decoding
637 else if (tiles_enabled_flag)
638 hvcc->parallelismType = 2; // tile-based parallel decoding
639 else
640 hvcc->parallelismType = 1; // slice-based parallel decoding
641
642 /* nothing useful for hvcC past this point */
643 return 0;
644}
645
646static uint8_t *nal_unit_extract_rbsp(const uint8_t *src, uint32_t src_len,
647 uint32_t *dst_len)
648{
649 uint8_t *dst;
650 uint32_t i, len;
651
652 dst = av_malloc(src_len + AV_INPUT_BUFFER_PADDING_SIZE);
653 if (!dst)
654 return NULL;
655
656 /* NAL unit header (2 bytes) */
657 i = len = 0;
658 while (i < 2 && i < src_len)
659 dst[len++] = src[i++];
660
661 while (i + 2 < src_len)
662 if (!src[i] && !src[i + 1] && src[i + 2] == 3) {
663 dst[len++] = src[i++];
664 dst[len++] = src[i++];
665 i++; // remove emulation_prevention_three_byte
666 } else
667 dst[len++] = src[i++];
668
669 while (i < src_len)
670 dst[len++] = src[i++];
671
672 *dst_len = len;
673 return dst;
674}
675
676
677
678static void nal_unit_parse_header(GetBitContext *gb, uint8_t *nal_type)
679{
680 skip_bits1(gb); // forbidden_zero_bit
681
682 *nal_type = get_bits(gb, 6);
683
684 /*
685 * nuh_layer_id u(6)
686 * nuh_temporal_id_plus1 u(3)
687 */
688 skip_bits(gb, 9);
689}
690
691static int hvcc_array_add_nal_unit(uint8_t *nal_buf, uint32_t nal_size,
692 uint8_t nal_type, int ps_array_completeness,
693 HEVCDecoderConfigurationRecord *hvcc)
694{
695 int ret;
696 uint8_t index;
697 uint16_t numNalus;
698 HVCCNALUnitArray *array;
699
700 for (index = 0; index < hvcc->numOfArrays; index++)
701 if (hvcc->array[index].NAL_unit_type == nal_type)
702 break;
703
704 if (index >= hvcc->numOfArrays) {
705 uint8_t i;
706
707 ret = av_reallocp_array(&hvcc->array, index + 1, sizeof(HVCCNALUnitArray));
708 if (ret < 0)
709 return ret;
710
711 for (i = hvcc->numOfArrays; i <= index; i++)
712 memset(&hvcc->array[i], 0, sizeof(HVCCNALUnitArray));
713 hvcc->numOfArrays = index + 1;
714 }
715
716 array = &hvcc->array[index];
717 numNalus = array->numNalus;
718
719 ret = av_reallocp_array(&array->nalUnit, numNalus + 1, sizeof(uint8_t*));
720 if (ret < 0)
721 return ret;
722
723 ret = av_reallocp_array(&array->nalUnitLength, numNalus + 1, sizeof(uint16_t));
724 if (ret < 0)
725 return ret;
726
727 array->nalUnit [numNalus] = nal_buf;
728 array->nalUnitLength[numNalus] = nal_size;
729 array->NAL_unit_type = nal_type;
730 array->numNalus++;
731
732 /*
733 * When the sample entry name is ‘hvc1’, the default and mandatory value of
734 * array_completeness is 1 for arrays of all types of parameter sets, and 0
735 * for all other arrays. When the sample entry name is ‘hev1’, the default
736 * value of array_completeness is 0 for all arrays.
737 */
738 if (nal_type == HEVC_NAL_VPS || nal_type == HEVC_NAL_SPS || nal_type == HEVC_NAL_PPS)
739 array->array_completeness = ps_array_completeness;
740
741 return 0;
742}
743
744static int hvcc_add_nal_unit(uint8_t *nal_buf, uint32_t nal_size,
745 int ps_array_completeness,
746 HEVCDecoderConfigurationRecord *hvcc)
747{
748 int ret = 0;
749 GetBitContext gbc;
750 uint8_t nal_type;
751 uint8_t *rbsp_buf;
752 uint32_t rbsp_size;
753
754 rbsp_buf = nal_unit_extract_rbsp(nal_buf, nal_size, &rbsp_size);
755 if (!rbsp_buf) {
756 ret = AVERROR(ENOMEM);
757 goto end;
758 }
759
760 ret = init_get_bits8(&gbc, rbsp_buf, rbsp_size);
761 if (ret < 0)
762 goto end;
763
764 nal_unit_parse_header(&gbc, &nal_type);
765
766 /*
767 * Note: only 'declarative' SEI messages are allowed in
768 * hvcC. Perhaps the SEI playload type should be checked
769 * and non-declarative SEI messages discarded?
770 */
771 switch (nal_type) {
772 case HEVC_NAL_VPS:
773 case HEVC_NAL_SPS:
774 case HEVC_NAL_PPS:
775 case HEVC_NAL_SEI_PREFIX:
776 case HEVC_NAL_SEI_SUFFIX:
777 ret = hvcc_array_add_nal_unit(nal_buf, nal_size, nal_type,
778 ps_array_completeness, hvcc);
779 if (ret < 0)
780 goto end;
781 else if (nal_type == HEVC_NAL_VPS)
782 ret = hvcc_parse_vps(&gbc, hvcc);
783 else if (nal_type == HEVC_NAL_SPS)
784 ret = hvcc_parse_sps(&gbc, hvcc);
785 else if (nal_type == HEVC_NAL_PPS)
786 ret = hvcc_parse_pps(&gbc, hvcc);
787 if (ret < 0)
788 goto end;
789 break;
790 default:
791 ret = AVERROR_INVALIDDATA;
792 goto end;
793 }
794
795end:
796 av_free(rbsp_buf);
797 return ret;
798}
799
800static void hvcc_init(HEVCDecoderConfigurationRecord *hvcc)
801{
802 memset(hvcc, 0, sizeof(HEVCDecoderConfigurationRecord));
803 hvcc->configurationVersion = 1;
804 hvcc->lengthSizeMinusOne = 3; // 4 bytes
805
806 /*
807 * The following fields have all their valid bits set by default,
808 * the ProfileTierLevel parsing code will unset them when needed.
809 */
810 hvcc->general_profile_compatibility_flags = 0xffffffff;
811 hvcc->general_constraint_indicator_flags = 0xffffffffffff;
812
813 /*
814 * Initialize this field with an invalid value which can be used to detect
815 * whether we didn't see any VUI (in which case it should be reset to zero).
816 */
817 hvcc->min_spatial_segmentation_idc = MAX_SPATIAL_SEGMENTATION + 1;
818}
819
820static void hvcc_close(HEVCDecoderConfigurationRecord *hvcc)
821{
822 uint8_t i;
823
824 for (i = 0; i < hvcc->numOfArrays; i++) {
825 hvcc->array[i].numNalus = 0;
826 av_freep(&hvcc->array[i].nalUnit);
827 av_freep(&hvcc->array[i].nalUnitLength);
828 }
829
830 hvcc->numOfArrays = 0;
831 av_freep(&hvcc->array);
832}
833
834static int hvcc_write(AVIOContext *pb, HEVCDecoderConfigurationRecord *hvcc)
835{
836 uint8_t i;
837 uint16_t j, vps_count = 0, sps_count = 0, pps_count = 0;
838
839 /*
840 * We only support writing HEVCDecoderConfigurationRecord version 1.
841 */
842 hvcc->configurationVersion = 1;
843
844 /*
845 * If min_spatial_segmentation_idc is invalid, reset to 0 (unspecified).
846 */
847 if (hvcc->min_spatial_segmentation_idc > MAX_SPATIAL_SEGMENTATION)
848 hvcc->min_spatial_segmentation_idc = 0;
849
850 /*
851 * parallelismType indicates the type of parallelism that is used to meet
852 * the restrictions imposed by min_spatial_segmentation_idc when the value
853 * of min_spatial_segmentation_idc is greater than 0.
854 */
855 if (!hvcc->min_spatial_segmentation_idc)
856 hvcc->parallelismType = 0;
857
858 /*
859 * It's unclear how to properly compute these fields, so
860 * let's always set them to values meaning 'unspecified'.
861 */
862 hvcc->avgFrameRate = 0;
863 hvcc->constantFrameRate = 0;
864
865 av_log(NULL, AV_LOG_TRACE, "configurationVersion: %"PRIu8"\n",
866 hvcc->configurationVersion);
867 av_log(NULL, AV_LOG_TRACE, "general_profile_space: %"PRIu8"\n",
868 hvcc->general_profile_space);
869 av_log(NULL, AV_LOG_TRACE, "general_tier_flag: %"PRIu8"\n",
870 hvcc->general_tier_flag);
871 av_log(NULL, AV_LOG_TRACE, "general_profile_idc: %"PRIu8"\n",
872 hvcc->general_profile_idc);
873 av_log(NULL, AV_LOG_TRACE, "general_profile_compatibility_flags: 0x%08"PRIx32"\n",
874 hvcc->general_profile_compatibility_flags);
875 av_log(NULL, AV_LOG_TRACE, "general_constraint_indicator_flags: 0x%012"PRIx64"\n",
876 hvcc->general_constraint_indicator_flags);
877 av_log(NULL, AV_LOG_TRACE, "general_level_idc: %"PRIu8"\n",
878 hvcc->general_level_idc);
879 av_log(NULL, AV_LOG_TRACE, "min_spatial_segmentation_idc: %"PRIu16"\n",
880 hvcc->min_spatial_segmentation_idc);
881 av_log(NULL, AV_LOG_TRACE, "parallelismType: %"PRIu8"\n",
882 hvcc->parallelismType);
883 av_log(NULL, AV_LOG_TRACE, "chromaFormat: %"PRIu8"\n",
884 hvcc->chromaFormat);
885 av_log(NULL, AV_LOG_TRACE, "bitDepthLumaMinus8: %"PRIu8"\n",
886 hvcc->bitDepthLumaMinus8);
887 av_log(NULL, AV_LOG_TRACE, "bitDepthChromaMinus8: %"PRIu8"\n",
888 hvcc->bitDepthChromaMinus8);
889 av_log(NULL, AV_LOG_TRACE, "avgFrameRate: %"PRIu16"\n",
890 hvcc->avgFrameRate);
891 av_log(NULL, AV_LOG_TRACE, "constantFrameRate: %"PRIu8"\n",
892 hvcc->constantFrameRate);
893 av_log(NULL, AV_LOG_TRACE, "numTemporalLayers: %"PRIu8"\n",
894 hvcc->numTemporalLayers);
895 av_log(NULL, AV_LOG_TRACE, "temporalIdNested: %"PRIu8"\n",
896 hvcc->temporalIdNested);
897 av_log(NULL, AV_LOG_TRACE, "lengthSizeMinusOne: %"PRIu8"\n",
898 hvcc->lengthSizeMinusOne);
899 av_log(NULL, AV_LOG_TRACE, "numOfArrays: %"PRIu8"\n",
900 hvcc->numOfArrays);
901 for (i = 0; i < hvcc->numOfArrays; i++) {
902 av_log(NULL, AV_LOG_TRACE, "array_completeness[%"PRIu8"]: %"PRIu8"\n",
903 i, hvcc->array[i].array_completeness);
904 av_log(NULL, AV_LOG_TRACE, "NAL_unit_type[%"PRIu8"]: %"PRIu8"\n",
905 i, hvcc->array[i].NAL_unit_type);
906 av_log(NULL, AV_LOG_TRACE, "numNalus[%"PRIu8"]: %"PRIu16"\n",
907 i, hvcc->array[i].numNalus);
908 for (j = 0; j < hvcc->array[i].numNalus; j++)
909 av_log(NULL, AV_LOG_TRACE,
910 "nalUnitLength[%"PRIu8"][%"PRIu16"]: %"PRIu16"\n",
911 i, j, hvcc->array[i].nalUnitLength[j]);
912 }
913
914 /*
915 * We need at least one of each: VPS, SPS and PPS.
916 */
917 for (i = 0; i < hvcc->numOfArrays; i++)
918 switch (hvcc->array[i].NAL_unit_type) {
919 case HEVC_NAL_VPS:
920 vps_count += hvcc->array[i].numNalus;
921 break;
922 case HEVC_NAL_SPS:
923 sps_count += hvcc->array[i].numNalus;
924 break;
925 case HEVC_NAL_PPS:
926 pps_count += hvcc->array[i].numNalus;
927 break;
928 default:
929 break;
930 }
931 if (!vps_count || vps_count > HEVC_MAX_VPS_COUNT ||
932 !sps_count || sps_count > HEVC_MAX_SPS_COUNT ||
933 !pps_count || pps_count > HEVC_MAX_PPS_COUNT)
934 return AVERROR_INVALIDDATA;
935
936 /* unsigned int(8) configurationVersion = 1; */
937 avio_w8(pb, hvcc->configurationVersion);
938
939 /*
940 * unsigned int(2) general_profile_space;
941 * unsigned int(1) general_tier_flag;
942 * unsigned int(5) general_profile_idc;
943 */
944 avio_w8(pb, hvcc->general_profile_space << 6 |
945 hvcc->general_tier_flag << 5 |
946 hvcc->general_profile_idc);
947
948 /* unsigned int(32) general_profile_compatibility_flags; */
949 avio_wb32(pb, hvcc->general_profile_compatibility_flags);
950
951 /* unsigned int(48) general_constraint_indicator_flags; */
952 avio_wb32(pb, hvcc->general_constraint_indicator_flags >> 16);
953 avio_wb16(pb, hvcc->general_constraint_indicator_flags);
954
955 /* unsigned int(8) general_level_idc; */
956 avio_w8(pb, hvcc->general_level_idc);
957
958 /*
959 * bit(4) reserved = ‘1111’b;
960 * unsigned int(12) min_spatial_segmentation_idc;
961 */
962 avio_wb16(pb, hvcc->min_spatial_segmentation_idc | 0xf000);
963
964 /*
965 * bit(6) reserved = ‘111111’b;
966 * unsigned int(2) parallelismType;
967 */
968 avio_w8(pb, hvcc->parallelismType | 0xfc);
969
970 /*
971 * bit(6) reserved = ‘111111’b;
972 * unsigned int(2) chromaFormat;
973 */
974 avio_w8(pb, hvcc->chromaFormat | 0xfc);
975
976 /*
977 * bit(5) reserved = ‘11111’b;
978 * unsigned int(3) bitDepthLumaMinus8;
979 */
980 avio_w8(pb, hvcc->bitDepthLumaMinus8 | 0xf8);
981
982 /*
983 * bit(5) reserved = ‘11111’b;
984 * unsigned int(3) bitDepthChromaMinus8;
985 */
986 avio_w8(pb, hvcc->bitDepthChromaMinus8 | 0xf8);
987
988 /* bit(16) avgFrameRate; */
989 avio_wb16(pb, hvcc->avgFrameRate);
990
991 /*
992 * bit(2) constantFrameRate;
993 * bit(3) numTemporalLayers;
994 * bit(1) temporalIdNested;
995 * unsigned int(2) lengthSizeMinusOne;
996 */
997 avio_w8(pb, hvcc->constantFrameRate << 6 |
998 hvcc->numTemporalLayers << 3 |
999 hvcc->temporalIdNested << 2 |
1000 hvcc->lengthSizeMinusOne);
1001
1002 /* unsigned int(8) numOfArrays; */
1003 avio_w8(pb, hvcc->numOfArrays);
1004
1005 for (i = 0; i < hvcc->numOfArrays; i++) {
1006 /*
1007 * bit(1) array_completeness;
1008 * unsigned int(1) reserved = 0;
1009 * unsigned int(6) NAL_unit_type;
1010 */
1011 avio_w8(pb, hvcc->array[i].array_completeness << 7 |
1012 hvcc->array[i].NAL_unit_type & 0x3f);
1013
1014 /* unsigned int(16) numNalus; */
1015 avio_wb16(pb, hvcc->array[i].numNalus);
1016
1017 for (j = 0; j < hvcc->array[i].numNalus; j++) {
1018 /* unsigned int(16) nalUnitLength; */
1019 avio_wb16(pb, hvcc->array[i].nalUnitLength[j]);
1020
1021 /* bit(8*nalUnitLength) nalUnit; */
1022 avio_write(pb, hvcc->array[i].nalUnit[j],
1023 hvcc->array[i].nalUnitLength[j]);
1024 }
1025 }
1026
1027 return 0;
1028}
1029
1030int ff_hevc_annexb2mp4(AVIOContext *pb, const uint8_t *buf_in,
1031 int size, int filter_ps, int *ps_count)
1032{
1033 int num_ps = 0, ret = 0;
1034 uint8_t *buf, *end, *start = NULL;
1035
1036 if (!filter_ps) {
1037 ret = ff_avc_parse_nal_units(pb, buf_in, size);
1038 goto end;
1039 }
1040
1041 ret = ff_avc_parse_nal_units_buf(buf_in, &start, &size);
1042 if (ret < 0)
1043 goto end;
1044
1045 ret = 0;
1046 buf = start;
1047 end = start + size;
1048
1049 while (end - buf > 4) {
1050 uint32_t len = FFMIN(AV_RB32(buf), end - buf - 4);
1051 uint8_t type = (buf[4] >> 1) & 0x3f;
1052
1053 buf += 4;
1054
1055 switch (type) {
1056 case HEVC_NAL_VPS:
1057 case HEVC_NAL_SPS:
1058 case HEVC_NAL_PPS:
1059 num_ps++;
1060 break;
1061 default:
1062 ret += 4 + len;
1063 avio_wb32(pb, len);
1064 avio_write(pb, buf, len);
1065 break;
1066 }
1067
1068 buf += len;
1069 }
1070
1071end:
1072 av_free(start);
1073 if (ps_count)
1074 *ps_count = num_ps;
1075 return ret;
1076}
1077
1078int ff_hevc_annexb2mp4_buf(const uint8_t *buf_in, uint8_t **buf_out,
1079 int *size, int filter_ps, int *ps_count)
1080{
1081 AVIOContext *pb;
1082 int ret;
1083
1084 ret = avio_open_dyn_buf(&pb);
1085 if (ret < 0)
1086 return ret;
1087
1088 ret = ff_hevc_annexb2mp4(pb, buf_in, *size, filter_ps, ps_count);
1089 *size = avio_close_dyn_buf(pb, buf_out);
1090
1091 return ret;
1092}
1093
1094int ff_isom_write_hvcc(AVIOContext *pb, const uint8_t *data,
1095 int size, int ps_array_completeness)
1096{
1097 int ret = 0;
1098 uint8_t *buf, *end, *start = NULL;
1099 HEVCDecoderConfigurationRecord hvcc;
1100
1101 hvcc_init(&hvcc);
1102
1103 if (size < 6) {
1104 /* We can't write a valid hvcC from the provided data */
1105 ret = AVERROR_INVALIDDATA;
1106 goto end;
1107 } else if (*data == 1) {
1108 /* Data is already hvcC-formatted */
1109 avio_write(pb, data, size);
1110 goto end;
1111 } else if (!(AV_RB24(data) == 1 || AV_RB32(data) == 1)) {
1112 /* Not a valid Annex B start code prefix */
1113 ret = AVERROR_INVALIDDATA;
1114 goto end;
1115 }
1116
1117 ret = ff_avc_parse_nal_units_buf(data, &start, &size);
1118 if (ret < 0)
1119 goto end;
1120
1121 buf = start;
1122 end = start + size;
1123
1124 while (end - buf > 4) {
1125 uint32_t len = FFMIN(AV_RB32(buf), end - buf - 4);
1126 uint8_t type = (buf[4] >> 1) & 0x3f;
1127
1128 buf += 4;
1129
1130 switch (type) {
1131 case HEVC_NAL_VPS:
1132 case HEVC_NAL_SPS:
1133 case HEVC_NAL_PPS:
1134 case HEVC_NAL_SEI_PREFIX:
1135 case HEVC_NAL_SEI_SUFFIX:
1136 ret = hvcc_add_nal_unit(buf, len, ps_array_completeness, &hvcc);
1137 if (ret < 0)
1138 goto end;
1139 break;
1140 default:
1141 break;
1142 }
1143
1144 buf += len;
1145 }
1146
1147 ret = hvcc_write(pb, &hvcc);
1148
1149end:
1150 hvcc_close(&hvcc);
1151 av_free(start);
1152 return ret;
1153}
1154
generated by cgit at 2024-05-18 01:39:03 (GMT)