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
31 #include "libavutil/internal.h"
36 #include "simple_idct.h"
37 #include "proresdec.h"
38 #include "proresdata.h"
40 static void permute(uint8_t *dst, const uint8_t *src, const uint8_t permutation[64])
43 for (i = 0; i < 64; i++)
44 dst[i] = permutation[src[i]];
47 static av_cold int decode_init(AVCodecContext *avctx)
49 ProresContext *ctx = avctx->priv_data;
50 uint8_t idct_permutation[64];
52 avctx->bits_per_raw_sample = 10;
54 ff_blockdsp_init(&ctx->bdsp, avctx);
55 ff_proresdsp_init(&ctx->prodsp, avctx);
57 ff_init_scantable_permutation(idct_permutation,
58 ctx->prodsp.idct_permutation_type);
60 permute(ctx->progressive_scan, ff_prores_progressive_scan, idct_permutation);
61 permute(ctx->interlaced_scan, ff_prores_interlaced_scan, idct_permutation);
66 static int decode_frame_header(ProresContext *ctx, const uint8_t *buf,
67 const int data_size, AVCodecContext *avctx)
69 int hdr_size, width, height, flags;
73 hdr_size = AV_RB16(buf);
74 ff_dlog(avctx, "header size %d\n", hdr_size);
75 if (hdr_size > data_size) {
76 av_log(avctx, AV_LOG_ERROR, "error, wrong header size\n");
77 return AVERROR_INVALIDDATA;
80 version = AV_RB16(buf + 2);
81 ff_dlog(avctx, "%.4s version %d\n", buf+4, version);
83 av_log(avctx, AV_LOG_ERROR, "unsupported version: %d\n", version);
84 return AVERROR_PATCHWELCOME;
87 width = AV_RB16(buf + 8);
88 height = AV_RB16(buf + 10);
89 if (width != avctx->width || height != avctx->height) {
90 av_log(avctx, AV_LOG_ERROR, "picture resolution change: %dx%d -> %dx%d\n",
91 avctx->width, avctx->height, width, height);
92 return AVERROR_PATCHWELCOME;
95 ctx->frame_type = (buf[12] >> 2) & 3;
96 ctx->alpha_info = buf[17] & 0xf;
98 if (ctx->alpha_info > 2) {
99 av_log(avctx, AV_LOG_ERROR, "Invalid alpha mode %d\n", ctx->alpha_info);
100 return AVERROR_INVALIDDATA;
102 if (avctx->skip_alpha) ctx->alpha_info = 0;
104 ff_dlog(avctx, "frame type %d\n", ctx->frame_type);
106 if (ctx->frame_type == 0) {
107 ctx->scan = ctx->progressive_scan; // permuted
109 ctx->scan = ctx->interlaced_scan; // permuted
110 ctx->frame->interlaced_frame = 1;
111 ctx->frame->top_field_first = ctx->frame_type == 1;
114 if (ctx->alpha_info) {
115 avctx->pix_fmt = (buf[12] & 0xC0) == 0xC0 ? AV_PIX_FMT_YUVA444P10 : AV_PIX_FMT_YUVA422P10;
117 avctx->pix_fmt = (buf[12] & 0xC0) == 0xC0 ? AV_PIX_FMT_YUV444P10 : AV_PIX_FMT_YUV422P10;
122 ff_dlog(avctx, "flags %x\n", flags);
125 if(buf + data_size - ptr < 64) {
126 av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
127 return AVERROR_INVALIDDATA;
129 permute(ctx->qmat_luma, ctx->prodsp.idct_permutation, ptr);
132 memset(ctx->qmat_luma, 4, 64);
136 if(buf + data_size - ptr < 64) {
137 av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
138 return AVERROR_INVALIDDATA;
140 permute(ctx->qmat_chroma, ctx->prodsp.idct_permutation, ptr);
142 memset(ctx->qmat_chroma, 4, 64);
148 static int decode_picture_header(AVCodecContext *avctx, const uint8_t *buf, const int buf_size)
150 ProresContext *ctx = avctx->priv_data;
151 int i, hdr_size, slice_count;
152 unsigned pic_data_size;
153 int log2_slice_mb_width, log2_slice_mb_height;
154 int slice_mb_count, mb_x, mb_y;
155 const uint8_t *data_ptr, *index_ptr;
157 hdr_size = buf[0] >> 3;
158 if (hdr_size < 8 || hdr_size > buf_size) {
159 av_log(avctx, AV_LOG_ERROR, "error, wrong picture header size\n");
160 return AVERROR_INVALIDDATA;
163 pic_data_size = AV_RB32(buf + 1);
164 if (pic_data_size > buf_size) {
165 av_log(avctx, AV_LOG_ERROR, "error, wrong picture data size\n");
166 return AVERROR_INVALIDDATA;
169 log2_slice_mb_width = buf[7] >> 4;
170 log2_slice_mb_height = buf[7] & 0xF;
171 if (log2_slice_mb_width > 3 || log2_slice_mb_height) {
172 av_log(avctx, AV_LOG_ERROR, "unsupported slice resolution: %dx%d\n",
173 1 << log2_slice_mb_width, 1 << log2_slice_mb_height);
174 return AVERROR_INVALIDDATA;
177 ctx->mb_width = (avctx->width + 15) >> 4;
179 ctx->mb_height = (avctx->height + 31) >> 5;
181 ctx->mb_height = (avctx->height + 15) >> 4;
183 // QT ignores the written value
184 // slice_count = AV_RB16(buf + 5);
185 slice_count = ctx->mb_height * ((ctx->mb_width >> log2_slice_mb_width) +
186 av_popcount(ctx->mb_width & (1 << log2_slice_mb_width) - 1));
188 if (ctx->slice_count != slice_count || !ctx->slices) {
189 av_freep(&ctx->slices);
190 ctx->slice_count = 0;
191 ctx->slices = av_mallocz_array(slice_count, sizeof(*ctx->slices));
193 return AVERROR(ENOMEM);
194 ctx->slice_count = slice_count;
198 return AVERROR(EINVAL);
200 if (hdr_size + slice_count*2 > buf_size) {
201 av_log(avctx, AV_LOG_ERROR, "error, wrong slice count\n");
202 return AVERROR_INVALIDDATA;
205 // parse slice information
206 index_ptr = buf + hdr_size;
207 data_ptr = index_ptr + slice_count*2;
209 slice_mb_count = 1 << log2_slice_mb_width;
213 for (i = 0; i < slice_count; i++) {
214 SliceContext *slice = &ctx->slices[i];
216 slice->data = data_ptr;
217 data_ptr += AV_RB16(index_ptr + i*2);
219 while (ctx->mb_width - mb_x < slice_mb_count)
220 slice_mb_count >>= 1;
224 slice->mb_count = slice_mb_count;
225 slice->data_size = data_ptr - slice->data;
227 if (slice->data_size < 6) {
228 av_log(avctx, AV_LOG_ERROR, "error, wrong slice data size\n");
229 return AVERROR_INVALIDDATA;
232 mb_x += slice_mb_count;
233 if (mb_x == ctx->mb_width) {
234 slice_mb_count = 1 << log2_slice_mb_width;
238 if (data_ptr > buf + buf_size) {
239 av_log(avctx, AV_LOG_ERROR, "error, slice out of bounds\n");
240 return AVERROR_INVALIDDATA;
244 if (mb_x || mb_y != ctx->mb_height) {
245 av_log(avctx, AV_LOG_ERROR, "error wrong mb count y %d h %d\n",
246 mb_y, ctx->mb_height);
247 return AVERROR_INVALIDDATA;
250 return pic_data_size;
253 #define DECODE_CODEWORD(val, codebook) \
255 unsigned int rice_order, exp_order, switch_bits; \
256 unsigned int q, buf, bits; \
258 UPDATE_CACHE(re, gb); \
259 buf = GET_CACHE(re, gb); \
261 /* number of bits to switch between rice and exp golomb */ \
262 switch_bits = codebook & 3; \
263 rice_order = codebook >> 5; \
264 exp_order = (codebook >> 2) & 7; \
266 q = 31 - av_log2(buf); \
268 if (q > switch_bits) { /* exp golomb */ \
269 bits = exp_order - switch_bits + (q<<1); \
270 val = SHOW_UBITS(re, gb, bits) - (1 << exp_order) + \
271 ((switch_bits + 1) << rice_order); \
272 SKIP_BITS(re, gb, bits); \
273 } else if (rice_order) { \
274 SKIP_BITS(re, gb, q+1); \
275 val = (q << rice_order) + SHOW_UBITS(re, gb, rice_order); \
276 SKIP_BITS(re, gb, rice_order); \
279 SKIP_BITS(re, gb, q+1); \
283 #define TOSIGNED(x) (((x) >> 1) ^ (-((x) & 1)))
285 #define FIRST_DC_CB 0xB8
287 static const uint8_t dc_codebook[7] = { 0x04, 0x28, 0x28, 0x4D, 0x4D, 0x70, 0x70};
289 static av_always_inline void decode_dc_coeffs(GetBitContext *gb, int16_t *out,
290 int blocks_per_slice)
297 DECODE_CODEWORD(code, FIRST_DC_CB);
298 prev_dc = TOSIGNED(code);
301 out += 64; // dc coeff for the next block
305 for (i = 1; i < blocks_per_slice; i++, out += 64) {
306 DECODE_CODEWORD(code, dc_codebook[FFMIN(code, 6U)]);
307 if(code) sign ^= -(code & 1);
309 prev_dc += (((code + 1) >> 1) ^ sign) - sign;
312 CLOSE_READER(re, gb);
315 // adaptive codebook switching lut according to previous run/level values
316 static const uint8_t run_to_cb[16] = { 0x06, 0x06, 0x05, 0x05, 0x04, 0x29, 0x29, 0x29, 0x29, 0x28, 0x28, 0x28, 0x28, 0x28, 0x28, 0x4C };
317 static const uint8_t lev_to_cb[10] = { 0x04, 0x0A, 0x05, 0x06, 0x04, 0x28, 0x28, 0x28, 0x28, 0x4C };
319 static av_always_inline int decode_ac_coeffs(AVCodecContext *avctx, GetBitContext *gb,
320 int16_t *out, int blocks_per_slice)
322 ProresContext *ctx = avctx->priv_data;
323 int block_mask, sign;
324 unsigned pos, run, level;
325 int max_coeffs, i, bits_left;
326 int log2_block_count = av_log2(blocks_per_slice);
329 UPDATE_CACHE(re, gb); \
333 max_coeffs = 64 << log2_block_count;
334 block_mask = blocks_per_slice - 1;
336 for (pos = block_mask;;) {
337 bits_left = gb->size_in_bits - re_index;
338 if (!bits_left || (bits_left < 32 && !SHOW_UBITS(re, gb, bits_left)))
341 DECODE_CODEWORD(run, run_to_cb[FFMIN(run, 15)]);
343 if (pos >= max_coeffs) {
344 av_log(avctx, AV_LOG_ERROR, "ac tex damaged %d, %d\n", pos, max_coeffs);
345 return AVERROR_INVALIDDATA;
348 DECODE_CODEWORD(level, lev_to_cb[FFMIN(level, 9)]);
351 i = pos >> log2_block_count;
353 sign = SHOW_SBITS(re, gb, 1);
354 SKIP_BITS(re, gb, 1);
355 out[((pos & block_mask) << 6) + ctx->scan[i]] = ((level ^ sign) - sign);
358 CLOSE_READER(re, gb);
362 static int decode_slice_luma(AVCodecContext *avctx, SliceContext *slice,
363 uint16_t *dst, int dst_stride,
364 const uint8_t *buf, unsigned buf_size,
367 ProresContext *ctx = avctx->priv_data;
368 LOCAL_ALIGNED_16(int16_t, blocks, [8*4*64]);
371 int i, blocks_per_slice = slice->mb_count<<2;
374 for (i = 0; i < blocks_per_slice; i++)
375 ctx->bdsp.clear_block(blocks+(i<<6));
377 init_get_bits(&gb, buf, buf_size << 3);
379 decode_dc_coeffs(&gb, blocks, blocks_per_slice);
380 if ((ret = decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice)) < 0)
384 for (i = 0; i < slice->mb_count; i++) {
385 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
386 ctx->prodsp.idct_put(dst +8, dst_stride, block+(1<<6), qmat);
387 ctx->prodsp.idct_put(dst+4*dst_stride , dst_stride, block+(2<<6), qmat);
388 ctx->prodsp.idct_put(dst+4*dst_stride+8, dst_stride, block+(3<<6), qmat);
395 static int decode_slice_chroma(AVCodecContext *avctx, SliceContext *slice,
396 uint16_t *dst, int dst_stride,
397 const uint8_t *buf, unsigned buf_size,
398 const int16_t *qmat, int log2_blocks_per_mb)
400 ProresContext *ctx = avctx->priv_data;
401 LOCAL_ALIGNED_16(int16_t, blocks, [8*4*64]);
404 int i, j, blocks_per_slice = slice->mb_count << log2_blocks_per_mb;
407 for (i = 0; i < blocks_per_slice; i++)
408 ctx->bdsp.clear_block(blocks+(i<<6));
410 init_get_bits(&gb, buf, buf_size << 3);
412 decode_dc_coeffs(&gb, blocks, blocks_per_slice);
413 if ((ret = decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice)) < 0)
417 for (i = 0; i < slice->mb_count; i++) {
418 for (j = 0; j < log2_blocks_per_mb; j++) {
419 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
420 ctx->prodsp.idct_put(dst+4*dst_stride, dst_stride, block+(1<<6), qmat);
428 static void unpack_alpha(GetBitContext *gb, uint16_t *dst, int num_coeffs,
431 const int mask = (1 << num_bits) - 1;
432 int i, idx, val, alpha_val;
439 val = get_bits(gb, num_bits);
442 val = get_bits(gb, num_bits == 16 ? 7 : 4);
444 val = (val + 2) >> 1;
448 alpha_val = (alpha_val + val) & mask;
449 if (num_bits == 16) {
450 dst[idx++] = alpha_val >> 6;
452 dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
454 if (idx >= num_coeffs)
456 } while (get_bits_left(gb)>0 && get_bits1(gb));
457 val = get_bits(gb, 4);
459 val = get_bits(gb, 11);
460 if (idx + val > num_coeffs)
461 val = num_coeffs - idx;
462 if (num_bits == 16) {
463 for (i = 0; i < val; i++)
464 dst[idx++] = alpha_val >> 6;
466 for (i = 0; i < val; i++)
467 dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
470 } while (idx < num_coeffs);
474 * Decode alpha slice plane.
476 static void decode_slice_alpha(ProresContext *ctx,
477 uint16_t *dst, int dst_stride,
478 const uint8_t *buf, int buf_size,
479 int blocks_per_slice)
483 LOCAL_ALIGNED_16(int16_t, blocks, [8*4*64]);
486 for (i = 0; i < blocks_per_slice<<2; i++)
487 ctx->bdsp.clear_block(blocks+(i<<6));
489 init_get_bits(&gb, buf, buf_size << 3);
491 if (ctx->alpha_info == 2) {
492 unpack_alpha(&gb, blocks, blocks_per_slice * 4 * 64, 16);
494 unpack_alpha(&gb, blocks, blocks_per_slice * 4 * 64, 8);
498 for (i = 0; i < 16; i++) {
499 memcpy(dst, block, 16 * blocks_per_slice * sizeof(*dst));
500 dst += dst_stride >> 1;
501 block += 16 * blocks_per_slice;
505 static int decode_slice_thread(AVCodecContext *avctx, void *arg, int jobnr, int threadnr)
507 ProresContext *ctx = avctx->priv_data;
508 SliceContext *slice = &ctx->slices[jobnr];
509 const uint8_t *buf = slice->data;
510 AVFrame *pic = ctx->frame;
511 int i, hdr_size, qscale, log2_chroma_blocks_per_mb;
512 int luma_stride, chroma_stride;
513 int y_data_size, u_data_size, v_data_size, a_data_size;
514 uint8_t *dest_y, *dest_u, *dest_v, *dest_a;
515 int16_t qmat_luma_scaled[64];
516 int16_t qmat_chroma_scaled[64];
521 //av_log(avctx, AV_LOG_INFO, "slice %d mb width %d mb x %d y %d\n",
522 // jobnr, slice->mb_count, slice->mb_x, slice->mb_y);
525 hdr_size = buf[0] >> 3;
526 qscale = av_clip(buf[1], 1, 224);
527 qscale = qscale > 128 ? qscale - 96 << 2: qscale;
528 y_data_size = AV_RB16(buf + 2);
529 u_data_size = AV_RB16(buf + 4);
530 v_data_size = slice->data_size - y_data_size - u_data_size - hdr_size;
531 if (hdr_size > 7) v_data_size = AV_RB16(buf + 6);
532 a_data_size = slice->data_size - y_data_size - u_data_size -
533 v_data_size - hdr_size;
535 if (y_data_size < 0 || u_data_size < 0 || v_data_size < 0
536 || hdr_size+y_data_size+u_data_size+v_data_size > slice->data_size){
537 av_log(avctx, AV_LOG_ERROR, "invalid plane data size\n");
538 return AVERROR_INVALIDDATA;
543 for (i = 0; i < 64; i++) {
544 qmat_luma_scaled [i] = ctx->qmat_luma [i] * qscale;
545 qmat_chroma_scaled[i] = ctx->qmat_chroma[i] * qscale;
548 if (ctx->frame_type == 0) {
549 luma_stride = pic->linesize[0];
550 chroma_stride = pic->linesize[1];
552 luma_stride = pic->linesize[0] << 1;
553 chroma_stride = pic->linesize[1] << 1;
556 if (avctx->pix_fmt == AV_PIX_FMT_YUV444P10 || avctx->pix_fmt == AV_PIX_FMT_YUVA444P10) {
558 log2_chroma_blocks_per_mb = 2;
561 log2_chroma_blocks_per_mb = 1;
564 dest_y = pic->data[0] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
565 dest_u = pic->data[1] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
566 dest_v = pic->data[2] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
567 dest_a = pic->data[3] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
569 if (ctx->frame_type && ctx->first_field ^ ctx->frame->top_field_first) {
570 dest_y += pic->linesize[0];
571 dest_u += pic->linesize[1];
572 dest_v += pic->linesize[2];
573 dest_a += pic->linesize[3];
576 ret = decode_slice_luma(avctx, slice, (uint16_t*)dest_y, luma_stride,
577 buf, y_data_size, qmat_luma_scaled);
581 if (!(avctx->flags & AV_CODEC_FLAG_GRAY) && (u_data_size + v_data_size) > 0) {
582 ret = decode_slice_chroma(avctx, slice, (uint16_t*)dest_u, chroma_stride,
583 buf + y_data_size, u_data_size,
584 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
588 ret = decode_slice_chroma(avctx, slice, (uint16_t*)dest_v, chroma_stride,
589 buf + y_data_size + u_data_size, v_data_size,
590 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
595 size_t mb_max_x = slice->mb_count << (mb_x_shift - 1);
596 for (size_t i = 0; i < 16; ++i)
597 for (size_t j = 0; j < mb_max_x; ++j) {
598 *(uint16_t*)(dest_u + (i * chroma_stride) + (j << 1)) = 511;
599 *(uint16_t*)(dest_v + (i * chroma_stride) + (j << 1)) = 511;
603 /* decode alpha plane if available */
604 if (ctx->alpha_info && pic->data[3] && a_data_size)
605 decode_slice_alpha(ctx, (uint16_t*)dest_a, luma_stride,
606 buf + y_data_size + u_data_size + v_data_size,
607 a_data_size, slice->mb_count);
613 static int decode_picture(AVCodecContext *avctx)
615 ProresContext *ctx = avctx->priv_data;
618 avctx->execute2(avctx, decode_slice_thread, NULL, NULL, ctx->slice_count);
620 for (i = 0; i < ctx->slice_count; i++)
621 if (ctx->slices[i].ret < 0)
622 return ctx->slices[i].ret;
627 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
630 ProresContext *ctx = avctx->priv_data;
631 AVFrame *frame = data;
632 const uint8_t *buf = avpkt->data;
633 int buf_size = avpkt->size;
634 int frame_hdr_size, pic_size, ret;
636 if (buf_size < 28 || AV_RL32(buf + 4) != AV_RL32("icpf")) {
637 av_log(avctx, AV_LOG_ERROR, "invalid frame header\n");
638 return AVERROR_INVALIDDATA;
642 ctx->frame->pict_type = AV_PICTURE_TYPE_I;
643 ctx->frame->key_frame = 1;
644 ctx->first_field = 1;
649 frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
650 if (frame_hdr_size < 0)
651 return frame_hdr_size;
653 buf += frame_hdr_size;
654 buf_size -= frame_hdr_size;
656 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
660 pic_size = decode_picture_header(avctx, buf, buf_size);
662 av_log(avctx, AV_LOG_ERROR, "error decoding picture header\n");
666 if ((ret = decode_picture(avctx)) < 0) {
667 av_log(avctx, AV_LOG_ERROR, "error decoding picture\n");
672 buf_size -= pic_size;
674 if (ctx->frame_type && buf_size > 0 && ctx->first_field) {
675 ctx->first_field = 0;
684 static av_cold int decode_close(AVCodecContext *avctx)
686 ProresContext *ctx = avctx->priv_data;
688 av_freep(&ctx->slices);
693 AVCodec ff_prores_decoder = {
695 .long_name = NULL_IF_CONFIG_SMALL("ProRes"),
696 .type = AVMEDIA_TYPE_VIDEO,
697 .id = AV_CODEC_ID_PRORES,
698 .priv_data_size = sizeof(ProresContext),
700 .close = decode_close,
701 .decode = decode_frame,
702 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_SLICE_THREADS,