2 * Copyright (c) 2010-2011 Maxim Poliakovski
3 * Copyright (c) 2010-2011 Elvis Presley
5 * This file is part of FFmpeg.
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.
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.
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
24 * Known FOURCCs: 'apch' (HQ), 'apcn' (SD), 'apcs' (LT), 'acpo' (Proxy), 'ap4h' (4444)
29 #define LONG_BITSTREAM_READER
34 #include "simple_idct.h"
35 #include "proresdec.h"
37 static void permute(uint8_t *dst, const uint8_t *src, const uint8_t permutation[64])
40 for (i = 0; i < 64; i++)
41 dst[i] = permutation[src[i]];
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
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,
66 static av_cold int decode_init(AVCodecContext *avctx)
68 ProresContext *ctx = avctx->priv_data;
69 uint8_t idct_permutation[64];
71 avctx->bits_per_raw_sample = 10;
73 ff_dsputil_init(&ctx->dsp, avctx);
74 ff_proresdsp_init(&ctx->prodsp, avctx);
76 ff_init_scantable_permutation(idct_permutation,
77 ctx->prodsp.idct_permutation_type);
79 permute(ctx->progressive_scan, progressive_scan, idct_permutation);
80 permute(ctx->interlaced_scan, interlaced_scan, idct_permutation);
85 static int decode_frame_header(ProresContext *ctx, const uint8_t *buf,
86 const int data_size, AVCodecContext *avctx)
88 int hdr_size, width, height, flags;
92 hdr_size = AV_RB16(buf);
93 av_dlog(avctx, "header size %d\n", hdr_size);
94 if (hdr_size > data_size) {
95 av_log(avctx, AV_LOG_ERROR, "error, wrong header size\n");
99 version = AV_RB16(buf + 2);
100 av_dlog(avctx, "%.4s version %d\n", buf+4, version);
102 av_log(avctx, AV_LOG_ERROR, "unsupported version: %d\n", version);
106 width = AV_RB16(buf + 8);
107 height = AV_RB16(buf + 10);
108 if (width != avctx->width || height != avctx->height) {
109 av_log(avctx, AV_LOG_ERROR, "picture resolution change: %dx%d -> %dx%d\n",
110 avctx->width, avctx->height, width, height);
114 ctx->frame_type = (buf[12] >> 2) & 3;
115 ctx->alpha_info = buf[17] & 0xf;
117 if (ctx->alpha_info > 2) {
118 av_log(avctx, AV_LOG_ERROR, "Invalid alpha mode %d\n", ctx->alpha_info);
119 return AVERROR_INVALIDDATA;
122 av_dlog(avctx, "frame type %d\n", ctx->frame_type);
124 if (ctx->frame_type == 0) {
125 ctx->scan = ctx->progressive_scan; // permuted
127 ctx->scan = ctx->interlaced_scan; // permuted
128 ctx->frame->interlaced_frame = 1;
129 ctx->frame->top_field_first = ctx->frame_type == 1;
132 if (ctx->alpha_info) {
133 avctx->pix_fmt = (buf[12] & 0xC0) == 0xC0 ? AV_PIX_FMT_YUVA444P10 : AV_PIX_FMT_YUVA422P10;
135 avctx->pix_fmt = (buf[12] & 0xC0) == 0xC0 ? AV_PIX_FMT_YUV444P10 : AV_PIX_FMT_YUV422P10;
140 av_dlog(avctx, "flags %x\n", flags);
143 if(buf + data_size - ptr < 64) {
144 av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
147 permute(ctx->qmat_luma, ctx->prodsp.idct_permutation, ptr);
150 memset(ctx->qmat_luma, 4, 64);
154 if(buf + data_size - ptr < 64) {
155 av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
158 permute(ctx->qmat_chroma, ctx->prodsp.idct_permutation, ptr);
160 memset(ctx->qmat_chroma, 4, 64);
166 static int decode_picture_header(AVCodecContext *avctx, const uint8_t *buf, const int buf_size)
168 ProresContext *ctx = avctx->priv_data;
169 int i, hdr_size, slice_count;
170 unsigned pic_data_size;
171 int log2_slice_mb_width, log2_slice_mb_height;
172 int slice_mb_count, mb_x, mb_y;
173 const uint8_t *data_ptr, *index_ptr;
175 hdr_size = buf[0] >> 3;
176 if (hdr_size < 8 || hdr_size > buf_size) {
177 av_log(avctx, AV_LOG_ERROR, "error, wrong picture header size\n");
181 pic_data_size = AV_RB32(buf + 1);
182 if (pic_data_size > buf_size) {
183 av_log(avctx, AV_LOG_ERROR, "error, wrong picture data size\n");
187 log2_slice_mb_width = buf[7] >> 4;
188 log2_slice_mb_height = buf[7] & 0xF;
189 if (log2_slice_mb_width > 3 || log2_slice_mb_height) {
190 av_log(avctx, AV_LOG_ERROR, "unsupported slice resolution: %dx%d\n",
191 1 << log2_slice_mb_width, 1 << log2_slice_mb_height);
195 ctx->mb_width = (avctx->width + 15) >> 4;
197 ctx->mb_height = (avctx->height + 31) >> 5;
199 ctx->mb_height = (avctx->height + 15) >> 4;
201 slice_count = AV_RB16(buf + 5);
203 if (ctx->slice_count != slice_count || !ctx->slices) {
204 av_freep(&ctx->slices);
205 ctx->slices = av_mallocz(slice_count * sizeof(*ctx->slices));
207 return AVERROR(ENOMEM);
208 ctx->slice_count = slice_count;
212 return AVERROR(EINVAL);
214 if (hdr_size + slice_count*2 > buf_size) {
215 av_log(avctx, AV_LOG_ERROR, "error, wrong slice count\n");
219 // parse slice information
220 index_ptr = buf + hdr_size;
221 data_ptr = index_ptr + slice_count*2;
223 slice_mb_count = 1 << log2_slice_mb_width;
227 for (i = 0; i < slice_count; i++) {
228 SliceContext *slice = &ctx->slices[i];
230 slice->data = data_ptr;
231 data_ptr += AV_RB16(index_ptr + i*2);
233 while (ctx->mb_width - mb_x < slice_mb_count)
234 slice_mb_count >>= 1;
238 slice->mb_count = slice_mb_count;
239 slice->data_size = data_ptr - slice->data;
241 if (slice->data_size < 6) {
242 av_log(avctx, AV_LOG_ERROR, "error, wrong slice data size\n");
246 mb_x += slice_mb_count;
247 if (mb_x == ctx->mb_width) {
248 slice_mb_count = 1 << log2_slice_mb_width;
252 if (data_ptr > buf + buf_size) {
253 av_log(avctx, AV_LOG_ERROR, "error, slice out of bounds\n");
258 if (mb_x || mb_y != ctx->mb_height) {
259 av_log(avctx, AV_LOG_ERROR, "error wrong mb count y %d h %d\n",
260 mb_y, ctx->mb_height);
264 return pic_data_size;
267 #define DECODE_CODEWORD(val, codebook) \
269 unsigned int rice_order, exp_order, switch_bits; \
270 unsigned int q, buf, bits; \
272 UPDATE_CACHE(re, gb); \
273 buf = GET_CACHE(re, gb); \
275 /* number of bits to switch between rice and exp golomb */ \
276 switch_bits = codebook & 3; \
277 rice_order = codebook >> 5; \
278 exp_order = (codebook >> 2) & 7; \
280 q = 31 - av_log2(buf); \
282 if (q > switch_bits) { /* exp golomb */ \
283 bits = exp_order - switch_bits + (q<<1); \
284 val = SHOW_UBITS(re, gb, bits) - (1 << exp_order) + \
285 ((switch_bits + 1) << rice_order); \
286 SKIP_BITS(re, gb, bits); \
287 } else if (rice_order) { \
288 SKIP_BITS(re, gb, q+1); \
289 val = (q << rice_order) + SHOW_UBITS(re, gb, rice_order); \
290 SKIP_BITS(re, gb, rice_order); \
293 SKIP_BITS(re, gb, q+1); \
297 #define TOSIGNED(x) (((x) >> 1) ^ (-((x) & 1)))
299 #define FIRST_DC_CB 0xB8
301 static const uint8_t dc_codebook[7] = { 0x04, 0x28, 0x28, 0x4D, 0x4D, 0x70, 0x70};
303 static av_always_inline void decode_dc_coeffs(GetBitContext *gb, int16_t *out,
304 int blocks_per_slice)
311 DECODE_CODEWORD(code, FIRST_DC_CB);
312 prev_dc = TOSIGNED(code);
315 out += 64; // dc coeff for the next block
319 for (i = 1; i < blocks_per_slice; i++, out += 64) {
320 DECODE_CODEWORD(code, dc_codebook[FFMIN(code, 6U)]);
321 if(code) sign ^= -(code & 1);
323 prev_dc += (((code + 1) >> 1) ^ sign) - sign;
326 CLOSE_READER(re, gb);
329 // adaptive codebook switching lut according to previous run/level values
330 static const uint8_t run_to_cb[16] = { 0x06, 0x06, 0x05, 0x05, 0x04, 0x29, 0x29, 0x29, 0x29, 0x28, 0x28, 0x28, 0x28, 0x28, 0x28, 0x4C };
331 static const uint8_t lev_to_cb[10] = { 0x04, 0x0A, 0x05, 0x06, 0x04, 0x28, 0x28, 0x28, 0x28, 0x4C };
333 static av_always_inline void decode_ac_coeffs(AVCodecContext *avctx, GetBitContext *gb,
334 int16_t *out, int blocks_per_slice)
336 ProresContext *ctx = avctx->priv_data;
337 int block_mask, sign;
338 unsigned pos, run, level;
339 int max_coeffs, i, bits_left;
340 int log2_block_count = av_log2(blocks_per_slice);
343 UPDATE_CACHE(re, gb); \
347 max_coeffs = 64 << log2_block_count;
348 block_mask = blocks_per_slice - 1;
350 for (pos = block_mask;;) {
351 bits_left = gb->size_in_bits - re_index;
352 if (!bits_left || (bits_left < 32 && !SHOW_UBITS(re, gb, bits_left)))
355 DECODE_CODEWORD(run, run_to_cb[FFMIN(run, 15)]);
357 if (pos >= max_coeffs) {
358 av_log(avctx, AV_LOG_ERROR, "ac tex damaged %d, %d\n", pos, max_coeffs);
362 DECODE_CODEWORD(level, lev_to_cb[FFMIN(level, 9)]);
365 i = pos >> log2_block_count;
367 sign = SHOW_SBITS(re, gb, 1);
368 SKIP_BITS(re, gb, 1);
369 out[((pos & block_mask) << 6) + ctx->scan[i]] = ((level ^ sign) - sign);
372 CLOSE_READER(re, gb);
375 static void decode_slice_luma(AVCodecContext *avctx, SliceContext *slice,
376 uint16_t *dst, int dst_stride,
377 const uint8_t *buf, unsigned buf_size,
380 ProresContext *ctx = avctx->priv_data;
381 LOCAL_ALIGNED_16(int16_t, blocks, [8*4*64]);
384 int i, blocks_per_slice = slice->mb_count<<2;
386 for (i = 0; i < blocks_per_slice; i++)
387 ctx->dsp.clear_block(blocks+(i<<6));
389 init_get_bits(&gb, buf, buf_size << 3);
391 decode_dc_coeffs(&gb, blocks, blocks_per_slice);
392 decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice);
395 for (i = 0; i < slice->mb_count; i++) {
396 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
397 ctx->prodsp.idct_put(dst +8, dst_stride, block+(1<<6), qmat);
398 ctx->prodsp.idct_put(dst+4*dst_stride , dst_stride, block+(2<<6), qmat);
399 ctx->prodsp.idct_put(dst+4*dst_stride+8, dst_stride, block+(3<<6), qmat);
405 static void decode_slice_chroma(AVCodecContext *avctx, SliceContext *slice,
406 uint16_t *dst, int dst_stride,
407 const uint8_t *buf, unsigned buf_size,
408 const int16_t *qmat, int log2_blocks_per_mb)
410 ProresContext *ctx = avctx->priv_data;
411 LOCAL_ALIGNED_16(int16_t, blocks, [8*4*64]);
414 int i, j, blocks_per_slice = slice->mb_count << log2_blocks_per_mb;
416 for (i = 0; i < blocks_per_slice; i++)
417 ctx->dsp.clear_block(blocks+(i<<6));
419 init_get_bits(&gb, buf, buf_size << 3);
421 decode_dc_coeffs(&gb, blocks, blocks_per_slice);
422 decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice);
425 for (i = 0; i < slice->mb_count; i++) {
426 for (j = 0; j < log2_blocks_per_mb; j++) {
427 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
428 ctx->prodsp.idct_put(dst+4*dst_stride, dst_stride, block+(1<<6), qmat);
435 static void unpack_alpha(GetBitContext *gb, uint16_t *dst, int num_coeffs,
438 const int mask = (1 << num_bits) - 1;
439 int i, idx, val, alpha_val;
446 val = get_bits(gb, num_bits);
449 val = get_bits(gb, num_bits == 16 ? 7 : 4);
451 val = (val + 2) >> 1;
455 alpha_val = (alpha_val + val) & mask;
456 if (num_bits == 16) {
457 dst[idx++] = alpha_val >> 6;
459 dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
461 if (idx >= num_coeffs)
463 } while (get_bits_left(gb)>0 && get_bits1(gb));
464 val = get_bits(gb, 4);
466 val = get_bits(gb, 11);
467 if (idx + val > num_coeffs)
468 val = num_coeffs - idx;
469 if (num_bits == 16) {
470 for (i = 0; i < val; i++)
471 dst[idx++] = alpha_val >> 6;
473 for (i = 0; i < val; i++)
474 dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
477 } while (idx < num_coeffs);
481 * Decode alpha slice plane.
483 static void decode_slice_alpha(ProresContext *ctx,
484 uint16_t *dst, int dst_stride,
485 const uint8_t *buf, int buf_size,
486 int blocks_per_slice)
490 LOCAL_ALIGNED_16(int16_t, blocks, [8*4*64]);
493 for (i = 0; i < blocks_per_slice<<2; i++)
494 ctx->dsp.clear_block(blocks+(i<<6));
496 init_get_bits(&gb, buf, buf_size << 3);
498 if (ctx->alpha_info == 2) {
499 unpack_alpha(&gb, blocks, blocks_per_slice * 4 * 64, 16);
501 unpack_alpha(&gb, blocks, blocks_per_slice * 4 * 64, 8);
505 for (i = 0; i < 16; i++) {
506 memcpy(dst, block, 16 * blocks_per_slice * sizeof(*dst));
507 dst += dst_stride >> 1;
508 block += 16 * blocks_per_slice;
512 static int decode_slice_thread(AVCodecContext *avctx, void *arg, int jobnr, int threadnr)
514 ProresContext *ctx = avctx->priv_data;
515 SliceContext *slice = &ctx->slices[jobnr];
516 const uint8_t *buf = slice->data;
517 AVFrame *pic = ctx->frame;
518 int i, hdr_size, qscale, log2_chroma_blocks_per_mb;
519 int luma_stride, chroma_stride;
520 int y_data_size, u_data_size, v_data_size, a_data_size;
521 uint8_t *dest_y, *dest_u, *dest_v, *dest_a;
522 int16_t qmat_luma_scaled[64];
523 int16_t qmat_chroma_scaled[64];
527 //av_log(avctx, AV_LOG_INFO, "slice %d mb width %d mb x %d y %d\n",
528 // jobnr, slice->mb_count, slice->mb_x, slice->mb_y);
531 hdr_size = buf[0] >> 3;
532 qscale = av_clip(buf[1], 1, 224);
533 qscale = qscale > 128 ? qscale - 96 << 2: qscale;
534 y_data_size = AV_RB16(buf + 2);
535 u_data_size = AV_RB16(buf + 4);
536 v_data_size = slice->data_size - y_data_size - u_data_size - hdr_size;
537 if (hdr_size > 7) v_data_size = AV_RB16(buf + 6);
538 a_data_size = slice->data_size - y_data_size - u_data_size -
539 v_data_size - hdr_size;
541 if (y_data_size < 0 || u_data_size < 0 || v_data_size < 0
542 || hdr_size+y_data_size+u_data_size+v_data_size > slice->data_size){
543 av_log(avctx, AV_LOG_ERROR, "invalid plane data size\n");
549 for (i = 0; i < 64; i++) {
550 qmat_luma_scaled [i] = ctx->qmat_luma [i] * qscale;
551 qmat_chroma_scaled[i] = ctx->qmat_chroma[i] * qscale;
554 if (ctx->frame_type == 0) {
555 luma_stride = pic->linesize[0];
556 chroma_stride = pic->linesize[1];
558 luma_stride = pic->linesize[0] << 1;
559 chroma_stride = pic->linesize[1] << 1;
562 if (avctx->pix_fmt == AV_PIX_FMT_YUV444P10 || avctx->pix_fmt == AV_PIX_FMT_YUVA444P10) {
564 log2_chroma_blocks_per_mb = 2;
567 log2_chroma_blocks_per_mb = 1;
570 dest_y = pic->data[0] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
571 dest_u = pic->data[1] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
572 dest_v = pic->data[2] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
573 dest_a = pic->data[3] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
575 if (ctx->frame_type && ctx->first_field ^ ctx->frame->top_field_first) {
576 dest_y += pic->linesize[0];
577 dest_u += pic->linesize[1];
578 dest_v += pic->linesize[2];
579 dest_a += pic->linesize[3];
582 decode_slice_luma(avctx, slice, (uint16_t*)dest_y, luma_stride,
583 buf, y_data_size, qmat_luma_scaled);
585 if (!(avctx->flags & CODEC_FLAG_GRAY)) {
586 decode_slice_chroma(avctx, slice, (uint16_t*)dest_u, chroma_stride,
587 buf + y_data_size, u_data_size,
588 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
589 decode_slice_chroma(avctx, slice, (uint16_t*)dest_v, chroma_stride,
590 buf + y_data_size + u_data_size, v_data_size,
591 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
593 /* decode alpha plane if available */
594 if (ctx->alpha_info && pic->data[3] && a_data_size)
595 decode_slice_alpha(ctx, (uint16_t*)dest_a, luma_stride,
596 buf + y_data_size + u_data_size + v_data_size,
597 a_data_size, slice->mb_count);
603 static int decode_picture(AVCodecContext *avctx)
605 ProresContext *ctx = avctx->priv_data;
608 avctx->execute2(avctx, decode_slice_thread, NULL, NULL, ctx->slice_count);
610 for (i = 0; i < ctx->slice_count; i++)
611 if (ctx->slices[i].ret < 0)
612 return ctx->slices[i].ret;
617 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
620 ProresContext *ctx = avctx->priv_data;
621 AVFrame *frame = data;
622 const uint8_t *buf = avpkt->data;
623 int buf_size = avpkt->size;
624 int frame_hdr_size, pic_size;
626 if (buf_size < 28 || AV_RL32(buf + 4) != AV_RL32("icpf")) {
627 av_log(avctx, AV_LOG_ERROR, "invalid frame header\n");
632 ctx->frame->pict_type = AV_PICTURE_TYPE_I;
633 ctx->frame->key_frame = 1;
634 ctx->first_field = 1;
639 frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
640 if (frame_hdr_size < 0)
643 buf += frame_hdr_size;
644 buf_size -= frame_hdr_size;
646 if (ff_get_buffer(avctx, frame, 0) < 0)
650 pic_size = decode_picture_header(avctx, buf, buf_size);
652 av_log(avctx, AV_LOG_ERROR, "error decoding picture header\n");
656 if (decode_picture(avctx)) {
657 av_log(avctx, AV_LOG_ERROR, "error decoding picture\n");
662 buf_size -= pic_size;
664 if (ctx->frame_type && buf_size > 0 && ctx->first_field) {
665 ctx->first_field = 0;
674 static av_cold int decode_close(AVCodecContext *avctx)
676 ProresContext *ctx = avctx->priv_data;
678 av_freep(&ctx->slices);
683 AVCodec ff_prores_decoder = {
685 .type = AVMEDIA_TYPE_VIDEO,
686 .id = AV_CODEC_ID_PRORES,
687 .priv_data_size = sizeof(ProresContext),
689 .close = decode_close,
690 .decode = decode_frame,
691 .long_name = NULL_IF_CONFIG_SMALL("ProRes"),
692 .capabilities = CODEC_CAP_DR1 | CODEC_CAP_SLICE_THREADS,