FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
proresdec2.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2010-2011 Maxim Poliakovski
3  * Copyright (c) 2010-2011 Elvis Presley
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  * Known FOURCCs: 'apch' (HQ), 'apcn' (SD), 'apcs' (LT), 'acpo' (Proxy), 'ap4h' (4444)
25  */
26 
27 //#define DEBUG
28 
29 #define LONG_BITSTREAM_READER
30 
31 #include "avcodec.h"
32 #include "get_bits.h"
33 #include "internal.h"
34 #include "simple_idct.h"
35 #include "proresdec.h"
36 
37 static void permute(uint8_t *dst, const uint8_t *src, const uint8_t permutation[64])
38 {
39  int i;
40  for (i = 0; i < 64; i++)
41  dst[i] = permutation[src[i]];
42 }
43 
44 static const uint8_t progressive_scan[64] = {
45  0, 1, 8, 9, 2, 3, 10, 11,
46  16, 17, 24, 25, 18, 19, 26, 27,
47  4, 5, 12, 20, 13, 6, 7, 14,
48  21, 28, 29, 22, 15, 23, 30, 31,
49  32, 33, 40, 48, 41, 34, 35, 42,
50  49, 56, 57, 50, 43, 36, 37, 44,
51  51, 58, 59, 52, 45, 38, 39, 46,
52  53, 60, 61, 54, 47, 55, 62, 63
53 };
54 
55 static const uint8_t interlaced_scan[64] = {
56  0, 8, 1, 9, 16, 24, 17, 25,
57  2, 10, 3, 11, 18, 26, 19, 27,
58  32, 40, 33, 34, 41, 48, 56, 49,
59  42, 35, 43, 50, 57, 58, 51, 59,
60  4, 12, 5, 6, 13, 20, 28, 21,
61  14, 7, 15, 22, 29, 36, 44, 37,
62  30, 23, 31, 38, 45, 52, 60, 53,
63  46, 39, 47, 54, 61, 62, 55, 63,
64 };
65 
67 {
68  ProresContext *ctx = avctx->priv_data;
69  uint8_t idct_permutation[64];
70 
71  avctx->bits_per_raw_sample = 10;
72 
73  ff_dsputil_init(&ctx->dsp, avctx);
74  ff_proresdsp_init(&ctx->prodsp, avctx);
75 
76  avctx->coded_frame = &ctx->frame;
78  ctx->frame.key_frame = 1;
79 
80  ff_init_scantable_permutation(idct_permutation,
82 
83  permute(ctx->progressive_scan, progressive_scan, idct_permutation);
84  permute(ctx->interlaced_scan, interlaced_scan, idct_permutation);
85 
86  return 0;
87 }
88 
89 static int decode_frame_header(ProresContext *ctx, const uint8_t *buf,
90  const int data_size, AVCodecContext *avctx)
91 {
92  int hdr_size, width, height, flags;
93  int version;
94  const uint8_t *ptr;
95 
96  hdr_size = AV_RB16(buf);
97  av_dlog(avctx, "header size %d\n", hdr_size);
98  if (hdr_size > data_size) {
99  av_log(avctx, AV_LOG_ERROR, "error, wrong header size\n");
100  return -1;
101  }
102 
103  version = AV_RB16(buf + 2);
104  av_dlog(avctx, "%.4s version %d\n", buf+4, version);
105  if (version > 1) {
106  av_log(avctx, AV_LOG_ERROR, "unsupported version: %d\n", version);
107  return -1;
108  }
109 
110  width = AV_RB16(buf + 8);
111  height = AV_RB16(buf + 10);
112  if (width != avctx->width || height != avctx->height) {
113  av_log(avctx, AV_LOG_ERROR, "picture resolution change: %dx%d -> %dx%d\n",
114  avctx->width, avctx->height, width, height);
115  return -1;
116  }
117 
118  ctx->frame_type = (buf[12] >> 2) & 3;
119 
120  av_dlog(avctx, "frame type %d\n", ctx->frame_type);
121 
122  if (ctx->frame_type == 0) {
123  ctx->scan = ctx->progressive_scan; // permuted
124  } else {
125  ctx->scan = ctx->interlaced_scan; // permuted
126  ctx->frame.interlaced_frame = 1;
127  ctx->frame.top_field_first = ctx->frame_type == 1;
128  }
129 
130  avctx->pix_fmt = (buf[12] & 0xC0) == 0xC0 ? AV_PIX_FMT_YUV444P10 : AV_PIX_FMT_YUV422P10;
131 
132  ptr = buf + 20;
133  flags = buf[19];
134  av_dlog(avctx, "flags %x\n", flags);
135 
136  if (flags & 2) {
137  if(buf + data_size - ptr < 64) {
138  av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
139  return -1;
140  }
141  permute(ctx->qmat_luma, ctx->prodsp.idct_permutation, ptr);
142  ptr += 64;
143  } else {
144  memset(ctx->qmat_luma, 4, 64);
145  }
146 
147  if (flags & 1) {
148  if(buf + data_size - ptr < 64) {
149  av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
150  return -1;
151  }
152  permute(ctx->qmat_chroma, ctx->prodsp.idct_permutation, ptr);
153  } else {
154  memset(ctx->qmat_chroma, 4, 64);
155  }
156 
157  return hdr_size;
158 }
159 
160 static int decode_picture_header(AVCodecContext *avctx, const uint8_t *buf, const int buf_size)
161 {
162  ProresContext *ctx = avctx->priv_data;
163  int i, hdr_size, slice_count;
164  unsigned pic_data_size;
165  int log2_slice_mb_width, log2_slice_mb_height;
166  int slice_mb_count, mb_x, mb_y;
167  const uint8_t *data_ptr, *index_ptr;
168 
169  hdr_size = buf[0] >> 3;
170  if (hdr_size < 8 || hdr_size > buf_size) {
171  av_log(avctx, AV_LOG_ERROR, "error, wrong picture header size\n");
172  return -1;
173  }
174 
175  pic_data_size = AV_RB32(buf + 1);
176  if (pic_data_size > buf_size) {
177  av_log(avctx, AV_LOG_ERROR, "error, wrong picture data size\n");
178  return -1;
179  }
180 
181  log2_slice_mb_width = buf[7] >> 4;
182  log2_slice_mb_height = buf[7] & 0xF;
183  if (log2_slice_mb_width > 3 || log2_slice_mb_height) {
184  av_log(avctx, AV_LOG_ERROR, "unsupported slice resolution: %dx%d\n",
185  1 << log2_slice_mb_width, 1 << log2_slice_mb_height);
186  return -1;
187  }
188 
189  ctx->mb_width = (avctx->width + 15) >> 4;
190  if (ctx->frame_type)
191  ctx->mb_height = (avctx->height + 31) >> 5;
192  else
193  ctx->mb_height = (avctx->height + 15) >> 4;
194 
195  slice_count = AV_RB16(buf + 5);
196 
197  if (ctx->slice_count != slice_count || !ctx->slices) {
198  av_freep(&ctx->slices);
199  ctx->slices = av_mallocz(slice_count * sizeof(*ctx->slices));
200  if (!ctx->slices)
201  return AVERROR(ENOMEM);
202  ctx->slice_count = slice_count;
203  }
204 
205  if (!slice_count)
206  return AVERROR(EINVAL);
207 
208  if (hdr_size + slice_count*2 > buf_size) {
209  av_log(avctx, AV_LOG_ERROR, "error, wrong slice count\n");
210  return -1;
211  }
212 
213  // parse slice information
214  index_ptr = buf + hdr_size;
215  data_ptr = index_ptr + slice_count*2;
216 
217  slice_mb_count = 1 << log2_slice_mb_width;
218  mb_x = 0;
219  mb_y = 0;
220 
221  for (i = 0; i < slice_count; i++) {
222  SliceContext *slice = &ctx->slices[i];
223 
224  slice->data = data_ptr;
225  data_ptr += AV_RB16(index_ptr + i*2);
226 
227  while (ctx->mb_width - mb_x < slice_mb_count)
228  slice_mb_count >>= 1;
229 
230  slice->mb_x = mb_x;
231  slice->mb_y = mb_y;
232  slice->mb_count = slice_mb_count;
233  slice->data_size = data_ptr - slice->data;
234 
235  if (slice->data_size < 6) {
236  av_log(avctx, AV_LOG_ERROR, "error, wrong slice data size\n");
237  return -1;
238  }
239 
240  mb_x += slice_mb_count;
241  if (mb_x == ctx->mb_width) {
242  slice_mb_count = 1 << log2_slice_mb_width;
243  mb_x = 0;
244  mb_y++;
245  }
246  if (data_ptr > buf + buf_size) {
247  av_log(avctx, AV_LOG_ERROR, "error, slice out of bounds\n");
248  return -1;
249  }
250  }
251 
252  if (mb_x || mb_y != ctx->mb_height) {
253  av_log(avctx, AV_LOG_ERROR, "error wrong mb count y %d h %d\n",
254  mb_y, ctx->mb_height);
255  return -1;
256  }
257 
258  return pic_data_size;
259 }
260 
261 #define DECODE_CODEWORD(val, codebook) \
262  do { \
263  unsigned int rice_order, exp_order, switch_bits; \
264  unsigned int q, buf, bits; \
265  \
266  UPDATE_CACHE(re, gb); \
267  buf = GET_CACHE(re, gb); \
268  \
269  /* number of bits to switch between rice and exp golomb */ \
270  switch_bits = codebook & 3; \
271  rice_order = codebook >> 5; \
272  exp_order = (codebook >> 2) & 7; \
273  \
274  q = 31 - av_log2(buf); \
275  \
276  if (q > switch_bits) { /* exp golomb */ \
277  bits = exp_order - switch_bits + (q<<1); \
278  val = SHOW_UBITS(re, gb, bits) - (1 << exp_order) + \
279  ((switch_bits + 1) << rice_order); \
280  SKIP_BITS(re, gb, bits); \
281  } else if (rice_order) { \
282  SKIP_BITS(re, gb, q+1); \
283  val = (q << rice_order) + SHOW_UBITS(re, gb, rice_order); \
284  SKIP_BITS(re, gb, rice_order); \
285  } else { \
286  val = q; \
287  SKIP_BITS(re, gb, q+1); \
288  } \
289  } while (0)
290 
291 #define TOSIGNED(x) (((x) >> 1) ^ (-((x) & 1)))
292 
293 #define FIRST_DC_CB 0xB8
294 
295 static const uint8_t dc_codebook[7] = { 0x04, 0x28, 0x28, 0x4D, 0x4D, 0x70, 0x70};
296 
298  int blocks_per_slice)
299 {
300  DCTELEM prev_dc;
301  int code, i, sign;
302 
303  OPEN_READER(re, gb);
304 
306  prev_dc = TOSIGNED(code);
307  out[0] = prev_dc;
308 
309  out += 64; // dc coeff for the next block
310 
311  code = 5;
312  sign = 0;
313  for (i = 1; i < blocks_per_slice; i++, out += 64) {
314  DECODE_CODEWORD(code, dc_codebook[FFMIN(code, 6U)]);
315  if(code) sign ^= -(code & 1);
316  else sign = 0;
317  prev_dc += (((code + 1) >> 1) ^ sign) - sign;
318  out[0] = prev_dc;
319  }
320  CLOSE_READER(re, gb);
321 }
322 
323 // adaptive codebook switching lut according to previous run/level values
324 static const uint8_t run_to_cb[16] = { 0x06, 0x06, 0x05, 0x05, 0x04, 0x29, 0x29, 0x29, 0x29, 0x28, 0x28, 0x28, 0x28, 0x28, 0x28, 0x4C };
325 static const uint8_t lev_to_cb[10] = { 0x04, 0x0A, 0x05, 0x06, 0x04, 0x28, 0x28, 0x28, 0x28, 0x4C };
326 
328  DCTELEM *out, int blocks_per_slice)
329 {
330  ProresContext *ctx = avctx->priv_data;
331  int block_mask, sign;
332  unsigned pos, run, level;
333  int max_coeffs, i, bits_left;
334  int log2_block_count = av_log2(blocks_per_slice);
335 
336  OPEN_READER(re, gb);
337  UPDATE_CACHE(re, gb); \
338  run = 4;
339  level = 2;
340 
341  max_coeffs = 64 << log2_block_count;
342  block_mask = blocks_per_slice - 1;
343 
344  for (pos = block_mask;;) {
345  bits_left = gb->size_in_bits - re_index;
346  if (!bits_left || (bits_left < 32 && !SHOW_UBITS(re, gb, bits_left)))
347  break;
348 
349  DECODE_CODEWORD(run, run_to_cb[FFMIN(run, 15)]);
350  pos += run + 1;
351  if (pos >= max_coeffs) {
352  av_log(avctx, AV_LOG_ERROR, "ac tex damaged %d, %d\n", pos, max_coeffs);
353  return;
354  }
355 
356  DECODE_CODEWORD(level, lev_to_cb[FFMIN(level, 9)]);
357  level += 1;
358 
359  i = pos >> log2_block_count;
360 
361  sign = SHOW_SBITS(re, gb, 1);
362  SKIP_BITS(re, gb, 1);
363  out[((pos & block_mask) << 6) + ctx->scan[i]] = ((level ^ sign) - sign);
364  }
365 
366  CLOSE_READER(re, gb);
367 }
368 
369 static void decode_slice_luma(AVCodecContext *avctx, SliceContext *slice,
370  uint16_t *dst, int dst_stride,
371  const uint8_t *buf, unsigned buf_size,
372  const int16_t *qmat)
373 {
374  ProresContext *ctx = avctx->priv_data;
375  LOCAL_ALIGNED_16(DCTELEM, blocks, [8*4*64]);
376  DCTELEM *block;
377  GetBitContext gb;
378  int i, blocks_per_slice = slice->mb_count<<2;
379 
380  for (i = 0; i < blocks_per_slice; i++)
381  ctx->dsp.clear_block(blocks+(i<<6));
382 
383  init_get_bits(&gb, buf, buf_size << 3);
384 
385  decode_dc_coeffs(&gb, blocks, blocks_per_slice);
386  decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice);
387 
388  block = blocks;
389  for (i = 0; i < slice->mb_count; i++) {
390  ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
391  ctx->prodsp.idct_put(dst +8, dst_stride, block+(1<<6), qmat);
392  ctx->prodsp.idct_put(dst+4*dst_stride , dst_stride, block+(2<<6), qmat);
393  ctx->prodsp.idct_put(dst+4*dst_stride+8, dst_stride, block+(3<<6), qmat);
394  block += 4*64;
395  dst += 16;
396  }
397 }
398 
400  uint16_t *dst, int dst_stride,
401  const uint8_t *buf, unsigned buf_size,
402  const int16_t *qmat, int log2_blocks_per_mb)
403 {
404  ProresContext *ctx = avctx->priv_data;
405  LOCAL_ALIGNED_16(DCTELEM, blocks, [8*4*64]);
406  DCTELEM *block;
407  GetBitContext gb;
408  int i, j, blocks_per_slice = slice->mb_count << log2_blocks_per_mb;
409 
410  for (i = 0; i < blocks_per_slice; i++)
411  ctx->dsp.clear_block(blocks+(i<<6));
412 
413  init_get_bits(&gb, buf, buf_size << 3);
414 
415  decode_dc_coeffs(&gb, blocks, blocks_per_slice);
416  decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice);
417 
418  block = blocks;
419  for (i = 0; i < slice->mb_count; i++) {
420  for (j = 0; j < log2_blocks_per_mb; j++) {
421  ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
422  ctx->prodsp.idct_put(dst+4*dst_stride, dst_stride, block+(1<<6), qmat);
423  block += 2*64;
424  dst += 8;
425  }
426  }
427 }
428 
429 static int decode_slice_thread(AVCodecContext *avctx, void *arg, int jobnr, int threadnr)
430 {
431  ProresContext *ctx = avctx->priv_data;
432  SliceContext *slice = &ctx->slices[jobnr];
433  const uint8_t *buf = slice->data;
434  AVFrame *pic = avctx->coded_frame;
435  int i, hdr_size, qscale, log2_chroma_blocks_per_mb;
436  int luma_stride, chroma_stride;
437  int y_data_size, u_data_size, v_data_size;
438  uint8_t *dest_y, *dest_u, *dest_v;
439  int16_t qmat_luma_scaled[64];
440  int16_t qmat_chroma_scaled[64];
441  int mb_x_shift;
442 
443  slice->ret = -1;
444  //av_log(avctx, AV_LOG_INFO, "slice %d mb width %d mb x %d y %d\n",
445  // jobnr, slice->mb_count, slice->mb_x, slice->mb_y);
446 
447  // slice header
448  hdr_size = buf[0] >> 3;
449  qscale = av_clip(buf[1], 1, 224);
450  qscale = qscale > 128 ? qscale - 96 << 2: qscale;
451  y_data_size = AV_RB16(buf + 2);
452  u_data_size = AV_RB16(buf + 4);
453  v_data_size = slice->data_size - y_data_size - u_data_size - hdr_size;
454  if (hdr_size > 7) v_data_size = AV_RB16(buf + 6);
455 
456  if (y_data_size < 0 || u_data_size < 0 || v_data_size < 0
457  || hdr_size+y_data_size+u_data_size+v_data_size > slice->data_size){
458  av_log(avctx, AV_LOG_ERROR, "invalid plane data size\n");
459  return -1;
460  }
461 
462  buf += hdr_size;
463 
464  for (i = 0; i < 64; i++) {
465  qmat_luma_scaled [i] = ctx->qmat_luma [i] * qscale;
466  qmat_chroma_scaled[i] = ctx->qmat_chroma[i] * qscale;
467  }
468 
469  if (ctx->frame_type == 0) {
470  luma_stride = pic->linesize[0];
471  chroma_stride = pic->linesize[1];
472  } else {
473  luma_stride = pic->linesize[0] << 1;
474  chroma_stride = pic->linesize[1] << 1;
475  }
476 
477  if (avctx->pix_fmt == AV_PIX_FMT_YUV444P10) {
478  mb_x_shift = 5;
479  log2_chroma_blocks_per_mb = 2;
480  } else {
481  mb_x_shift = 4;
482  log2_chroma_blocks_per_mb = 1;
483  }
484 
485  dest_y = pic->data[0] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
486  dest_u = pic->data[1] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
487  dest_v = pic->data[2] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
488 
489  if (ctx->frame_type && ctx->first_field ^ ctx->frame.top_field_first) {
490  dest_y += pic->linesize[0];
491  dest_u += pic->linesize[1];
492  dest_v += pic->linesize[2];
493  }
494 
495  decode_slice_luma(avctx, slice, (uint16_t*)dest_y, luma_stride,
496  buf, y_data_size, qmat_luma_scaled);
497 
498  if (!(avctx->flags & CODEC_FLAG_GRAY)) {
499  decode_slice_chroma(avctx, slice, (uint16_t*)dest_u, chroma_stride,
500  buf + y_data_size, u_data_size,
501  qmat_chroma_scaled, log2_chroma_blocks_per_mb);
502  decode_slice_chroma(avctx, slice, (uint16_t*)dest_v, chroma_stride,
503  buf + y_data_size + u_data_size, v_data_size,
504  qmat_chroma_scaled, log2_chroma_blocks_per_mb);
505  }
506 
507  slice->ret = 0;
508  return 0;
509 }
510 
511 static int decode_picture(AVCodecContext *avctx)
512 {
513  ProresContext *ctx = avctx->priv_data;
514  int i;
515 
516  avctx->execute2(avctx, decode_slice_thread, NULL, NULL, ctx->slice_count);
517 
518  for (i = 0; i < ctx->slice_count; i++)
519  if (ctx->slices[i].ret < 0)
520  return ctx->slices[i].ret;
521 
522  return 0;
523 }
524 
525 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
526  AVPacket *avpkt)
527 {
528  ProresContext *ctx = avctx->priv_data;
529  AVFrame *frame = avctx->coded_frame;
530  const uint8_t *buf = avpkt->data;
531  int buf_size = avpkt->size;
532  int frame_hdr_size, pic_size;
533 
534  if (buf_size < 28 || AV_RL32(buf + 4) != AV_RL32("icpf")) {
535  av_log(avctx, AV_LOG_ERROR, "invalid frame header\n");
536  return -1;
537  }
538 
539  ctx->first_field = 1;
540 
541  buf += 8;
542  buf_size -= 8;
543 
544  frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
545  if (frame_hdr_size < 0)
546  return -1;
547 
548  buf += frame_hdr_size;
549  buf_size -= frame_hdr_size;
550 
551  if (frame->data[0])
552  avctx->release_buffer(avctx, frame);
553 
554  if (ff_get_buffer(avctx, frame) < 0)
555  return -1;
556 
558  pic_size = decode_picture_header(avctx, buf, buf_size);
559  if (pic_size < 0) {
560  av_log(avctx, AV_LOG_ERROR, "error decoding picture header\n");
561  return -1;
562  }
563 
564  if (decode_picture(avctx)) {
565  av_log(avctx, AV_LOG_ERROR, "error decoding picture\n");
566  return -1;
567  }
568 
569  buf += pic_size;
570  buf_size -= pic_size;
571 
572  if (ctx->frame_type && buf_size > 0 && ctx->first_field) {
573  ctx->first_field = 0;
574  goto decode_picture;
575  }
576 
577  *got_frame = 1;
578  *(AVFrame*)data = *frame;
579 
580  return avpkt->size;
581 }
582 
584 {
585  ProresContext *ctx = avctx->priv_data;
586 
587  AVFrame *frame = avctx->coded_frame;
588  if (frame->data[0])
589  avctx->release_buffer(avctx, frame);
590  av_freep(&ctx->slices);
591 
592  return 0;
593 }
594 
596  .name = "prores",
597  .type = AVMEDIA_TYPE_VIDEO,
598  .id = AV_CODEC_ID_PRORES,
599  .priv_data_size = sizeof(ProresContext),
600  .init = decode_init,
601  .close = decode_close,
602  .decode = decode_frame,
603  .long_name = NULL_IF_CONFIG_SMALL("ProRes"),
604  .capabilities = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS,
605 };