3 * Copyright (c) 2011 Konstantin Shishkov
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
30 #include "libavutil/intreadwrite.h"
33 #include "bytestream.h"
38 static int build_huff10(const uint8_t *src, VLC *vlc, int *fsym)
49 for (i = 0; i < 1024; i++) {
53 qsort(he, 1024, sizeof(*he), ff_ut10_huff_cmp_len);
61 while (he[last].len == 255 && last)
64 if (he[last].len > 32) {
69 for (i = last; i >= 0; i--) {
70 codes[i] = code >> (32 - he[i].len);
73 code += 0x80000000u >> (he[i].len - 1);
76 return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 11), last + 1,
77 bits, sizeof(*bits), sizeof(*bits),
78 codes, sizeof(*codes), sizeof(*codes),
79 syms, sizeof(*syms), sizeof(*syms), 0);
82 static int build_huff(const uint8_t *src, VLC *vlc, int *fsym)
93 for (i = 0; i < 256; i++) {
97 qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);
105 while (he[last].len == 255 && last)
108 if (he[last].len > 32)
112 for (i = last; i >= 0; i--) {
113 codes[i] = code >> (32 - he[i].len);
116 code += 0x80000000u >> (he[i].len - 1);
119 return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 11), last + 1,
120 bits, sizeof(*bits), sizeof(*bits),
121 codes, sizeof(*codes), sizeof(*codes),
122 syms, sizeof(*syms), sizeof(*syms), 0);
125 static int decode_plane10(UtvideoContext *c, int plane_no,
126 uint16_t *dst, int step, int stride,
127 int width, int height,
128 const uint8_t *src, const uint8_t *huff,
131 int i, j, slice, pix, ret;
137 if ((ret = build_huff10(huff, &vlc, &fsym)) < 0) {
138 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
141 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
143 for (slice = 0; slice < c->slices; slice++) {
147 send = (height * (slice + 1) / c->slices);
148 dest = dst + sstart * stride;
151 for (j = sstart; j < send; j++) {
152 for (i = 0; i < width * step; i += step) {
167 for (slice = 0; slice < c->slices; slice++) {
169 int slice_data_start, slice_data_end, slice_size;
172 send = (height * (slice + 1) / c->slices);
173 dest = dst + sstart * stride;
175 // slice offset and size validation was done earlier
176 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
177 slice_data_end = AV_RL32(src + slice * 4);
178 slice_size = slice_data_end - slice_data_start;
181 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
182 "yet a slice has a length of zero.\n");
186 memcpy(c->slice_bits, src + slice_data_start + c->slices * 4,
188 memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
189 c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
190 (uint32_t *) c->slice_bits,
191 (slice_data_end - slice_data_start + 3) >> 2);
192 init_get_bits(&gb, c->slice_bits, slice_size * 8);
195 for (j = sstart; j < send; j++) {
196 for (i = 0; i < width * step; i += step) {
197 if (get_bits_left(&gb) <= 0) {
198 av_log(c->avctx, AV_LOG_ERROR,
199 "Slice decoding ran out of bits\n");
202 pix = get_vlc2(&gb, vlc.table, vlc.bits, 3);
204 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
216 if (get_bits_left(&gb) > 32)
217 av_log(c->avctx, AV_LOG_WARNING,
218 "%d bits left after decoding slice\n", get_bits_left(&gb));
226 return AVERROR_INVALIDDATA;
229 static int decode_plane(UtvideoContext *c, int plane_no,
230 uint8_t *dst, int step, int stride,
231 int width, int height,
232 const uint8_t *src, int use_pred)
234 int i, j, slice, pix;
239 const int cmask = ~(!plane_no && c->avctx->pix_fmt == AV_PIX_FMT_YUV420P);
241 if (build_huff(src, &vlc, &fsym)) {
242 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
243 return AVERROR_INVALIDDATA;
245 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
247 for (slice = 0; slice < c->slices; slice++) {
251 send = (height * (slice + 1) / c->slices) & cmask;
252 dest = dst + sstart * stride;
255 for (j = sstart; j < send; j++) {
256 for (i = 0; i < width * step; i += step) {
273 for (slice = 0; slice < c->slices; slice++) {
275 int slice_data_start, slice_data_end, slice_size;
278 send = (height * (slice + 1) / c->slices) & cmask;
279 dest = dst + sstart * stride;
281 // slice offset and size validation was done earlier
282 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
283 slice_data_end = AV_RL32(src + slice * 4);
284 slice_size = slice_data_end - slice_data_start;
287 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
288 "yet a slice has a length of zero.\n");
292 memcpy(c->slice_bits, src + slice_data_start + c->slices * 4,
294 memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
295 c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
296 (uint32_t *) c->slice_bits,
297 (slice_data_end - slice_data_start + 3) >> 2);
298 init_get_bits(&gb, c->slice_bits, slice_size * 8);
301 for (j = sstart; j < send; j++) {
302 for (i = 0; i < width * step; i += step) {
303 if (get_bits_left(&gb) <= 0) {
304 av_log(c->avctx, AV_LOG_ERROR,
305 "Slice decoding ran out of bits\n");
308 pix = get_vlc2(&gb, vlc.table, vlc.bits, 3);
310 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
321 if (get_bits_left(&gb) > 32)
322 av_log(c->avctx, AV_LOG_WARNING,
323 "%d bits left after decoding slice\n", get_bits_left(&gb));
331 return AVERROR_INVALIDDATA;
334 static void restore_rgb_planes(uint8_t *src, int step, int stride, int width,
340 for (j = 0; j < height; j++) {
341 for (i = 0; i < width * step; i += step) {
345 src[i] = r + g - 0x80;
346 src[i + 2] = b + g - 0x80;
352 static void restore_median(uint8_t *src, int step, int stride,
353 int width, int height, int slices, int rmode)
358 int slice_start, slice_height;
359 const int cmask = ~rmode;
361 for (slice = 0; slice < slices; slice++) {
362 slice_start = ((slice * height) / slices) & cmask;
363 slice_height = ((((slice + 1) * height) / slices) & cmask) -
368 bsrc = src + slice_start * stride;
370 // first line - left neighbour prediction
373 for (i = step; i < width * step; i += step) {
378 if (slice_height <= 1)
380 // second line - first element has top prediction, the rest uses median
384 for (i = step; i < width * step; i += step) {
385 B = bsrc[i - stride];
386 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
391 // the rest of lines use continuous median prediction
392 for (j = 2; j < slice_height; j++) {
393 for (i = 0; i < width * step; i += step) {
394 B = bsrc[i - stride];
395 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
404 /* UtVideo interlaced mode treats every two lines as a single one,
405 * so restoring function should take care of possible padding between
406 * two parts of the same "line".
408 static void restore_median_il(uint8_t *src, int step, int stride,
409 int width, int height, int slices, int rmode)
414 int slice_start, slice_height;
415 const int cmask = ~(rmode ? 3 : 1);
416 const int stride2 = stride << 1;
418 for (slice = 0; slice < slices; slice++) {
419 slice_start = ((slice * height) / slices) & cmask;
420 slice_height = ((((slice + 1) * height) / slices) & cmask) -
426 bsrc = src + slice_start * stride;
428 // first line - left neighbour prediction
431 for (i = step; i < width * step; i += step) {
435 for (i = 0; i < width * step; i += step) {
436 bsrc[stride + i] += A;
437 A = bsrc[stride + i];
440 if (slice_height <= 1)
442 // second line - first element has top prediction, the rest uses median
446 for (i = step; i < width * step; i += step) {
447 B = bsrc[i - stride2];
448 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
452 for (i = 0; i < width * step; i += step) {
453 B = bsrc[i - stride];
454 bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C));
456 A = bsrc[stride + i];
459 // the rest of lines use continuous median prediction
460 for (j = 2; j < slice_height; j++) {
461 for (i = 0; i < width * step; i += step) {
462 B = bsrc[i - stride2];
463 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
467 for (i = 0; i < width * step; i += step) {
468 B = bsrc[i - stride];
469 bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C));
471 A = bsrc[i + stride];
478 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
481 const uint8_t *buf = avpkt->data;
482 int buf_size = avpkt->size;
483 UtvideoContext *c = avctx->priv_data;
485 const uint8_t *plane_start[5];
486 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
489 ThreadFrame frame = { .f = data };
491 if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0)
494 /* parse plane structure to get frame flags and validate slice offsets */
495 bytestream2_init(&gb, buf, buf_size);
497 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
498 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
499 return AVERROR_INVALIDDATA;
501 c->frame_info = bytestream2_get_le32u(&gb);
502 c->slices = ((c->frame_info >> 16) & 0xff) + 1;
503 for (i = 0; i < c->planes; i++) {
504 plane_start[i] = gb.buffer;
505 if (bytestream2_get_bytes_left(&gb) < 1024 + 4 * c->slices) {
506 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
507 return AVERROR_INVALIDDATA;
511 for (j = 0; j < c->slices; j++) {
512 slice_end = bytestream2_get_le32u(&gb);
513 if (slice_end < 0 || slice_end < slice_start ||
514 bytestream2_get_bytes_left(&gb) < slice_end) {
515 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
516 return AVERROR_INVALIDDATA;
518 slice_size = slice_end - slice_start;
519 slice_start = slice_end;
520 max_slice_size = FFMAX(max_slice_size, slice_size);
522 plane_size = slice_end;
523 bytestream2_skipu(&gb, plane_size);
524 bytestream2_skipu(&gb, 1024);
526 plane_start[c->planes] = gb.buffer;
528 for (i = 0; i < c->planes; i++) {
529 plane_start[i] = gb.buffer;
530 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
531 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
532 return AVERROR_INVALIDDATA;
534 bytestream2_skipu(&gb, 256);
537 for (j = 0; j < c->slices; j++) {
538 slice_end = bytestream2_get_le32u(&gb);
539 if (slice_end < 0 || slice_end < slice_start ||
540 bytestream2_get_bytes_left(&gb) < slice_end) {
541 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
542 return AVERROR_INVALIDDATA;
544 slice_size = slice_end - slice_start;
545 slice_start = slice_end;
546 max_slice_size = FFMAX(max_slice_size, slice_size);
548 plane_size = slice_end;
549 bytestream2_skipu(&gb, plane_size);
551 plane_start[c->planes] = gb.buffer;
552 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
553 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
554 return AVERROR_INVALIDDATA;
556 c->frame_info = bytestream2_get_le32u(&gb);
558 av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
561 c->frame_pred = (c->frame_info >> 8) & 3;
563 if (c->frame_pred == PRED_GRADIENT) {
564 avpriv_request_sample(avctx, "Frame with gradient prediction");
565 return AVERROR_PATCHWELCOME;
568 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
569 max_slice_size + AV_INPUT_BUFFER_PADDING_SIZE);
571 if (!c->slice_bits) {
572 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
573 return AVERROR(ENOMEM);
576 switch (c->avctx->pix_fmt) {
577 case AV_PIX_FMT_RGB24:
578 case AV_PIX_FMT_RGBA:
579 for (i = 0; i < c->planes; i++) {
580 ret = decode_plane(c, i, frame.f->data[0] + ff_ut_rgb_order[i],
581 c->planes, frame.f->linesize[0], avctx->width,
582 avctx->height, plane_start[i],
583 c->frame_pred == PRED_LEFT);
586 if (c->frame_pred == PRED_MEDIAN) {
587 if (!c->interlaced) {
588 restore_median(frame.f->data[0] + ff_ut_rgb_order[i],
589 c->planes, frame.f->linesize[0], avctx->width,
590 avctx->height, c->slices, 0);
592 restore_median_il(frame.f->data[0] + ff_ut_rgb_order[i],
593 c->planes, frame.f->linesize[0],
594 avctx->width, avctx->height, c->slices,
599 restore_rgb_planes(frame.f->data[0], c->planes, frame.f->linesize[0],
600 avctx->width, avctx->height);
602 case AV_PIX_FMT_YUV420P:
603 for (i = 0; i < 3; i++) {
604 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
605 avctx->width >> !!i, avctx->height >> !!i,
606 plane_start[i], c->frame_pred == PRED_LEFT);
609 if (c->frame_pred == PRED_MEDIAN) {
610 if (!c->interlaced) {
611 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
612 avctx->width >> !!i, avctx->height >> !!i,
615 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
617 avctx->height >> !!i,
623 case AV_PIX_FMT_YUV422P:
624 for (i = 0; i < 3; i++) {
625 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
626 avctx->width >> !!i, avctx->height,
627 plane_start[i], c->frame_pred == PRED_LEFT);
630 if (c->frame_pred == PRED_MEDIAN) {
631 if (!c->interlaced) {
632 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
633 avctx->width >> !!i, avctx->height,
636 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
637 avctx->width >> !!i, avctx->height,
643 case AV_PIX_FMT_YUV422P10:
644 for (i = 0; i < 3; i++) {
645 ret = decode_plane10(c, i, (uint16_t *)frame.f->data[i], 1, frame.f->linesize[i] / 2,
646 avctx->width >> !!i, avctx->height,
647 plane_start[i], plane_start[i + 1] - 1024, c->frame_pred == PRED_LEFT);
654 frame.f->key_frame = 1;
655 frame.f->pict_type = AV_PICTURE_TYPE_I;
656 frame.f->interlaced_frame = !!c->interlaced;
660 /* always report that the buffer was completely consumed */
664 static av_cold int decode_init(AVCodecContext *avctx)
666 UtvideoContext * const c = avctx->priv_data;
670 ff_bswapdsp_init(&c->bdsp);
672 if (avctx->extradata_size >= 16) {
673 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
674 avctx->extradata[3], avctx->extradata[2],
675 avctx->extradata[1], avctx->extradata[0]);
676 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
677 AV_RB32(avctx->extradata + 4));
678 c->frame_info_size = AV_RL32(avctx->extradata + 8);
679 c->flags = AV_RL32(avctx->extradata + 12);
681 if (c->frame_info_size != 4)
682 avpriv_request_sample(avctx, "Frame info not 4 bytes");
683 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
684 c->slices = (c->flags >> 24) + 1;
685 c->compression = c->flags & 1;
686 c->interlaced = c->flags & 0x800;
687 } else if (avctx->extradata_size == 8) {
688 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
689 avctx->extradata[3], avctx->extradata[2],
690 avctx->extradata[1], avctx->extradata[0]);
691 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
692 AV_RB32(avctx->extradata + 4));
695 c->frame_info_size = 4;
697 av_log(avctx, AV_LOG_ERROR,
698 "Insufficient extradata size %d, should be at least 16\n",
699 avctx->extradata_size);
700 return AVERROR_INVALIDDATA;
703 c->slice_bits_size = 0;
705 switch (avctx->codec_tag) {
706 case MKTAG('U', 'L', 'R', 'G'):
708 avctx->pix_fmt = AV_PIX_FMT_RGB24;
710 case MKTAG('U', 'L', 'R', 'A'):
712 avctx->pix_fmt = AV_PIX_FMT_RGBA;
714 case MKTAG('U', 'L', 'Y', '0'):
716 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
717 avctx->colorspace = AVCOL_SPC_BT470BG;
719 case MKTAG('U', 'L', 'Y', '2'):
721 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
722 avctx->colorspace = AVCOL_SPC_BT470BG;
724 case MKTAG('U', 'Q', 'Y', '2'):
726 avctx->pix_fmt = AV_PIX_FMT_YUV422P10;
728 case MKTAG('U', 'L', 'H', '0'):
730 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
731 avctx->colorspace = AVCOL_SPC_BT709;
733 case MKTAG('U', 'L', 'H', '2'):
735 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
736 avctx->colorspace = AVCOL_SPC_BT709;
739 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
741 return AVERROR_INVALIDDATA;
747 static av_cold int decode_end(AVCodecContext *avctx)
749 UtvideoContext * const c = avctx->priv_data;
751 av_freep(&c->slice_bits);
756 AVCodec ff_utvideo_decoder = {
758 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
759 .type = AVMEDIA_TYPE_VIDEO,
760 .id = AV_CODEC_ID_UTVIDEO,
761 .priv_data_size = sizeof(UtvideoContext),
764 .decode = decode_frame,
765 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,