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;
120 avctx->color_primaries = buf[14];
121 avctx->color_trc = buf[15];
122 avctx->colorspace = buf[16];
123 avctx->color_range = AVCOL_RANGE_MPEG;
127 ff_dlog(avctx, "flags %x\n", flags);
130 if(buf + data_size - ptr < 64) {
131 av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
132 return AVERROR_INVALIDDATA;
134 permute(ctx->qmat_luma, ctx->prodsp.idct_permutation, ptr);
137 memset(ctx->qmat_luma, 4, 64);
141 if(buf + data_size - ptr < 64) {
142 av_log(avctx, AV_LOG_ERROR, "Header truncated\n");
143 return AVERROR_INVALIDDATA;
145 permute(ctx->qmat_chroma, ctx->prodsp.idct_permutation, ptr);
147 memset(ctx->qmat_chroma, 4, 64);
153 static int decode_picture_header(AVCodecContext *avctx, const uint8_t *buf, const int buf_size)
155 ProresContext *ctx = avctx->priv_data;
156 int i, hdr_size, slice_count;
157 unsigned pic_data_size;
158 int log2_slice_mb_width, log2_slice_mb_height;
159 int slice_mb_count, mb_x, mb_y;
160 const uint8_t *data_ptr, *index_ptr;
162 hdr_size = buf[0] >> 3;
163 if (hdr_size < 8 || hdr_size > buf_size) {
164 av_log(avctx, AV_LOG_ERROR, "error, wrong picture header size\n");
165 return AVERROR_INVALIDDATA;
168 pic_data_size = AV_RB32(buf + 1);
169 if (pic_data_size > buf_size) {
170 av_log(avctx, AV_LOG_ERROR, "error, wrong picture data size\n");
171 return AVERROR_INVALIDDATA;
174 log2_slice_mb_width = buf[7] >> 4;
175 log2_slice_mb_height = buf[7] & 0xF;
176 if (log2_slice_mb_width > 3 || log2_slice_mb_height) {
177 av_log(avctx, AV_LOG_ERROR, "unsupported slice resolution: %dx%d\n",
178 1 << log2_slice_mb_width, 1 << log2_slice_mb_height);
179 return AVERROR_INVALIDDATA;
182 ctx->mb_width = (avctx->width + 15) >> 4;
184 ctx->mb_height = (avctx->height + 31) >> 5;
186 ctx->mb_height = (avctx->height + 15) >> 4;
188 // QT ignores the written value
189 // slice_count = AV_RB16(buf + 5);
190 slice_count = ctx->mb_height * ((ctx->mb_width >> log2_slice_mb_width) +
191 av_popcount(ctx->mb_width & (1 << log2_slice_mb_width) - 1));
193 if (ctx->slice_count != slice_count || !ctx->slices) {
194 av_freep(&ctx->slices);
195 ctx->slice_count = 0;
196 ctx->slices = av_mallocz_array(slice_count, sizeof(*ctx->slices));
198 return AVERROR(ENOMEM);
199 ctx->slice_count = slice_count;
203 return AVERROR(EINVAL);
205 if (hdr_size + slice_count*2 > buf_size) {
206 av_log(avctx, AV_LOG_ERROR, "error, wrong slice count\n");
207 return AVERROR_INVALIDDATA;
210 // parse slice information
211 index_ptr = buf + hdr_size;
212 data_ptr = index_ptr + slice_count*2;
214 slice_mb_count = 1 << log2_slice_mb_width;
218 for (i = 0; i < slice_count; i++) {
219 SliceContext *slice = &ctx->slices[i];
221 slice->data = data_ptr;
222 data_ptr += AV_RB16(index_ptr + i*2);
224 while (ctx->mb_width - mb_x < slice_mb_count)
225 slice_mb_count >>= 1;
229 slice->mb_count = slice_mb_count;
230 slice->data_size = data_ptr - slice->data;
232 if (slice->data_size < 6) {
233 av_log(avctx, AV_LOG_ERROR, "error, wrong slice data size\n");
234 return AVERROR_INVALIDDATA;
237 mb_x += slice_mb_count;
238 if (mb_x == ctx->mb_width) {
239 slice_mb_count = 1 << log2_slice_mb_width;
243 if (data_ptr > buf + buf_size) {
244 av_log(avctx, AV_LOG_ERROR, "error, slice out of bounds\n");
245 return AVERROR_INVALIDDATA;
249 if (mb_x || mb_y != ctx->mb_height) {
250 av_log(avctx, AV_LOG_ERROR, "error wrong mb count y %d h %d\n",
251 mb_y, ctx->mb_height);
252 return AVERROR_INVALIDDATA;
255 return pic_data_size;
258 #define DECODE_CODEWORD(val, codebook, SKIP) \
260 unsigned int rice_order, exp_order, switch_bits; \
261 unsigned int q, buf, bits; \
263 UPDATE_CACHE(re, gb); \
264 buf = GET_CACHE(re, gb); \
266 /* number of bits to switch between rice and exp golomb */ \
267 switch_bits = codebook & 3; \
268 rice_order = codebook >> 5; \
269 exp_order = (codebook >> 2) & 7; \
271 q = 31 - av_log2(buf); \
273 if (q > switch_bits) { /* exp golomb */ \
274 bits = exp_order - switch_bits + (q<<1); \
275 if (bits > FFMIN(MIN_CACHE_BITS, 31)) \
276 return AVERROR_INVALIDDATA; \
277 val = SHOW_UBITS(re, gb, bits) - (1 << exp_order) + \
278 ((switch_bits + 1) << rice_order); \
279 SKIP(re, gb, bits); \
280 } else if (rice_order) { \
281 SKIP_BITS(re, gb, q+1); \
282 val = (q << rice_order) + SHOW_UBITS(re, gb, rice_order); \
283 SKIP(re, gb, rice_order); \
290 #define TOSIGNED(x) (((x) >> 1) ^ (-((x) & 1)))
292 #define FIRST_DC_CB 0xB8
294 static const uint8_t dc_codebook[7] = { 0x04, 0x28, 0x28, 0x4D, 0x4D, 0x70, 0x70};
296 static av_always_inline int decode_dc_coeffs(GetBitContext *gb, int16_t *out,
297 int blocks_per_slice)
304 DECODE_CODEWORD(code, FIRST_DC_CB, LAST_SKIP_BITS);
305 prev_dc = TOSIGNED(code);
308 out += 64; // dc coeff for the next block
312 for (i = 1; i < blocks_per_slice; i++, out += 64) {
313 DECODE_CODEWORD(code, dc_codebook[FFMIN(code, 6U)], LAST_SKIP_BITS);
314 if(code) sign ^= -(code & 1);
316 prev_dc += (((code + 1) >> 1) ^ sign) - sign;
319 CLOSE_READER(re, gb);
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 };
327 static av_always_inline int decode_ac_coeffs(AVCodecContext *avctx, GetBitContext *gb,
328 int16_t *out, int blocks_per_slice)
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);
337 UPDATE_CACHE(re, gb); \
341 max_coeffs = 64 << log2_block_count;
342 block_mask = blocks_per_slice - 1;
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)))
349 DECODE_CODEWORD(run, run_to_cb[FFMIN(run, 15)], LAST_SKIP_BITS);
351 if (pos >= max_coeffs) {
352 av_log(avctx, AV_LOG_ERROR, "ac tex damaged %d, %d\n", pos, max_coeffs);
353 return AVERROR_INVALIDDATA;
356 DECODE_CODEWORD(level, lev_to_cb[FFMIN(level, 9)], SKIP_BITS);
359 i = pos >> log2_block_count;
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);
366 CLOSE_READER(re, gb);
370 static int decode_slice_luma(AVCodecContext *avctx, SliceContext *slice,
371 uint16_t *dst, int dst_stride,
372 const uint8_t *buf, unsigned buf_size,
375 ProresContext *ctx = avctx->priv_data;
376 LOCAL_ALIGNED_32(int16_t, blocks, [8*4*64]);
379 int i, blocks_per_slice = slice->mb_count<<2;
382 for (i = 0; i < blocks_per_slice; i++)
383 ctx->bdsp.clear_block(blocks+(i<<6));
385 init_get_bits(&gb, buf, buf_size << 3);
387 if ((ret = decode_dc_coeffs(&gb, blocks, blocks_per_slice)) < 0)
389 if ((ret = decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice)) < 0)
393 for (i = 0; i < slice->mb_count; i++) {
394 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
395 ctx->prodsp.idct_put(dst +8, dst_stride, block+(1<<6), qmat);
396 ctx->prodsp.idct_put(dst+4*dst_stride , dst_stride, block+(2<<6), qmat);
397 ctx->prodsp.idct_put(dst+4*dst_stride+8, dst_stride, block+(3<<6), qmat);
404 static int decode_slice_chroma(AVCodecContext *avctx, SliceContext *slice,
405 uint16_t *dst, int dst_stride,
406 const uint8_t *buf, unsigned buf_size,
407 const int16_t *qmat, int log2_blocks_per_mb)
409 ProresContext *ctx = avctx->priv_data;
410 LOCAL_ALIGNED_32(int16_t, blocks, [8*4*64]);
413 int i, j, blocks_per_slice = slice->mb_count << log2_blocks_per_mb;
416 for (i = 0; i < blocks_per_slice; i++)
417 ctx->bdsp.clear_block(blocks+(i<<6));
419 init_get_bits(&gb, buf, buf_size << 3);
421 if ((ret = decode_dc_coeffs(&gb, blocks, blocks_per_slice)) < 0)
423 if ((ret = decode_ac_coeffs(avctx, &gb, blocks, blocks_per_slice)) < 0)
427 for (i = 0; i < slice->mb_count; i++) {
428 for (j = 0; j < log2_blocks_per_mb; j++) {
429 ctx->prodsp.idct_put(dst, dst_stride, block+(0<<6), qmat);
430 ctx->prodsp.idct_put(dst+4*dst_stride, dst_stride, block+(1<<6), qmat);
438 static void unpack_alpha(GetBitContext *gb, uint16_t *dst, int num_coeffs,
441 const int mask = (1 << num_bits) - 1;
442 int i, idx, val, alpha_val;
449 val = get_bits(gb, num_bits);
452 val = get_bits(gb, num_bits == 16 ? 7 : 4);
454 val = (val + 2) >> 1;
458 alpha_val = (alpha_val + val) & mask;
459 if (num_bits == 16) {
460 dst[idx++] = alpha_val >> 6;
462 dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
464 if (idx >= num_coeffs)
466 } while (get_bits_left(gb)>0 && get_bits1(gb));
467 val = get_bits(gb, 4);
469 val = get_bits(gb, 11);
470 if (idx + val > num_coeffs)
471 val = num_coeffs - idx;
472 if (num_bits == 16) {
473 for (i = 0; i < val; i++)
474 dst[idx++] = alpha_val >> 6;
476 for (i = 0; i < val; i++)
477 dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
480 } while (idx < num_coeffs);
484 * Decode alpha slice plane.
486 static void decode_slice_alpha(ProresContext *ctx,
487 uint16_t *dst, int dst_stride,
488 const uint8_t *buf, int buf_size,
489 int blocks_per_slice)
493 LOCAL_ALIGNED_32(int16_t, blocks, [8*4*64]);
496 for (i = 0; i < blocks_per_slice<<2; i++)
497 ctx->bdsp.clear_block(blocks+(i<<6));
499 init_get_bits(&gb, buf, buf_size << 3);
501 if (ctx->alpha_info == 2) {
502 unpack_alpha(&gb, blocks, blocks_per_slice * 4 * 64, 16);
504 unpack_alpha(&gb, blocks, blocks_per_slice * 4 * 64, 8);
508 for (i = 0; i < 16; i++) {
509 memcpy(dst, block, 16 * blocks_per_slice * sizeof(*dst));
510 dst += dst_stride >> 1;
511 block += 16 * blocks_per_slice;
515 static int decode_slice_thread(AVCodecContext *avctx, void *arg, int jobnr, int threadnr)
517 ProresContext *ctx = avctx->priv_data;
518 SliceContext *slice = &ctx->slices[jobnr];
519 const uint8_t *buf = slice->data;
520 AVFrame *pic = ctx->frame;
521 int i, hdr_size, qscale, log2_chroma_blocks_per_mb;
522 int luma_stride, chroma_stride;
523 int y_data_size, u_data_size, v_data_size, a_data_size;
524 uint8_t *dest_y, *dest_u, *dest_v, *dest_a;
525 LOCAL_ALIGNED_16(int16_t, qmat_luma_scaled, [64]);
526 LOCAL_ALIGNED_16(int16_t, qmat_chroma_scaled,[64]);
531 //av_log(avctx, AV_LOG_INFO, "slice %d mb width %d mb x %d y %d\n",
532 // jobnr, slice->mb_count, slice->mb_x, slice->mb_y);
535 hdr_size = buf[0] >> 3;
536 qscale = av_clip(buf[1], 1, 224);
537 qscale = qscale > 128 ? qscale - 96 << 2: qscale;
538 y_data_size = AV_RB16(buf + 2);
539 u_data_size = AV_RB16(buf + 4);
540 v_data_size = slice->data_size - y_data_size - u_data_size - hdr_size;
541 if (hdr_size > 7) v_data_size = AV_RB16(buf + 6);
542 a_data_size = slice->data_size - y_data_size - u_data_size -
543 v_data_size - hdr_size;
545 if (y_data_size < 0 || u_data_size < 0 || v_data_size < 0
546 || hdr_size+y_data_size+u_data_size+v_data_size > slice->data_size){
547 av_log(avctx, AV_LOG_ERROR, "invalid plane data size\n");
548 return AVERROR_INVALIDDATA;
553 for (i = 0; i < 64; i++) {
554 qmat_luma_scaled [i] = ctx->qmat_luma [i] * qscale;
555 qmat_chroma_scaled[i] = ctx->qmat_chroma[i] * qscale;
558 if (ctx->frame_type == 0) {
559 luma_stride = pic->linesize[0];
560 chroma_stride = pic->linesize[1];
562 luma_stride = pic->linesize[0] << 1;
563 chroma_stride = pic->linesize[1] << 1;
566 if (avctx->pix_fmt == AV_PIX_FMT_YUV444P10 || avctx->pix_fmt == AV_PIX_FMT_YUVA444P10) {
568 log2_chroma_blocks_per_mb = 2;
571 log2_chroma_blocks_per_mb = 1;
574 dest_y = pic->data[0] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
575 dest_u = pic->data[1] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
576 dest_v = pic->data[2] + (slice->mb_y << 4) * chroma_stride + (slice->mb_x << mb_x_shift);
577 dest_a = pic->data[3] + (slice->mb_y << 4) * luma_stride + (slice->mb_x << 5);
579 if (ctx->frame_type && ctx->first_field ^ ctx->frame->top_field_first) {
580 dest_y += pic->linesize[0];
581 dest_u += pic->linesize[1];
582 dest_v += pic->linesize[2];
583 dest_a += pic->linesize[3];
586 ret = decode_slice_luma(avctx, slice, (uint16_t*)dest_y, luma_stride,
587 buf, y_data_size, qmat_luma_scaled);
591 if (!(avctx->flags & AV_CODEC_FLAG_GRAY) && (u_data_size + v_data_size) > 0) {
592 ret = decode_slice_chroma(avctx, slice, (uint16_t*)dest_u, chroma_stride,
593 buf + y_data_size, u_data_size,
594 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
598 ret = decode_slice_chroma(avctx, slice, (uint16_t*)dest_v, chroma_stride,
599 buf + y_data_size + u_data_size, v_data_size,
600 qmat_chroma_scaled, log2_chroma_blocks_per_mb);
605 size_t mb_max_x = slice->mb_count << (mb_x_shift - 1);
607 for (i = 0; i < 16; ++i)
608 for (j = 0; j < mb_max_x; ++j) {
609 *(uint16_t*)(dest_u + (i * chroma_stride) + (j << 1)) = 511;
610 *(uint16_t*)(dest_v + (i * chroma_stride) + (j << 1)) = 511;
614 /* decode alpha plane if available */
615 if (ctx->alpha_info && pic->data[3] && a_data_size)
616 decode_slice_alpha(ctx, (uint16_t*)dest_a, luma_stride,
617 buf + y_data_size + u_data_size + v_data_size,
618 a_data_size, slice->mb_count);
624 static int decode_picture(AVCodecContext *avctx)
626 ProresContext *ctx = avctx->priv_data;
630 avctx->execute2(avctx, decode_slice_thread, NULL, NULL, ctx->slice_count);
632 for (i = 0; i < ctx->slice_count; i++)
633 error += ctx->slices[i].ret < 0;
636 ctx->frame->decode_error_flags = FF_DECODE_ERROR_INVALID_BITSTREAM;
637 if (error < ctx->slice_count)
640 return ctx->slices[0].ret;
643 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
646 ProresContext *ctx = avctx->priv_data;
647 AVFrame *frame = data;
648 const uint8_t *buf = avpkt->data;
649 int buf_size = avpkt->size;
650 int frame_hdr_size, pic_size, ret;
652 if (buf_size < 28 || AV_RL32(buf + 4) != AV_RL32("icpf")) {
653 av_log(avctx, AV_LOG_ERROR, "invalid frame header\n");
654 return AVERROR_INVALIDDATA;
658 ctx->frame->pict_type = AV_PICTURE_TYPE_I;
659 ctx->frame->key_frame = 1;
660 ctx->first_field = 1;
665 frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
666 if (frame_hdr_size < 0)
667 return frame_hdr_size;
669 buf += frame_hdr_size;
670 buf_size -= frame_hdr_size;
672 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
676 pic_size = decode_picture_header(avctx, buf, buf_size);
678 av_log(avctx, AV_LOG_ERROR, "error decoding picture header\n");
682 if ((ret = decode_picture(avctx)) < 0) {
683 av_log(avctx, AV_LOG_ERROR, "error decoding picture\n");
688 buf_size -= pic_size;
690 if (ctx->frame_type && buf_size > 0 && ctx->first_field) {
691 ctx->first_field = 0;
700 static av_cold int decode_close(AVCodecContext *avctx)
702 ProresContext *ctx = avctx->priv_data;
704 av_freep(&ctx->slices);
709 AVCodec ff_prores_decoder = {
711 .long_name = NULL_IF_CONFIG_SMALL("ProRes (iCodec Pro)"),
712 .type = AVMEDIA_TYPE_VIDEO,
713 .id = AV_CODEC_ID_PRORES,
714 .priv_data_size = sizeof(ProresContext),
716 .close = decode_close,
717 .decode = decode_frame,
718 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_SLICE_THREADS,