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_huff(const uint8_t *src, VLC *vlc, int *fsym)
49 for (i = 0; i < 256; i++) {
53 qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);
63 while (he[last].len == 255 && last)
67 for (i = last; i >= 0; i--) {
68 codes[i] = code >> (32 - he[i].len);
71 code += 0x80000000u >> (he[i].len - 1);
74 return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 11), last + 1,
75 bits, sizeof(*bits), sizeof(*bits),
76 codes, sizeof(*codes), sizeof(*codes),
77 syms, sizeof(*syms), sizeof(*syms), 0);
80 static int decode_plane(UtvideoContext *c, int plane_no,
81 uint8_t *dst, int step, int stride,
82 int width, int height,
83 const uint8_t *src, int use_pred)
90 const int cmask = ~(!plane_no && c->avctx->pix_fmt == AV_PIX_FMT_YUV420P);
92 if (build_huff(src, &vlc, &fsym)) {
93 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
94 return AVERROR_INVALIDDATA;
96 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
98 for (slice = 0; slice < c->slices; slice++) {
102 send = (height * (slice + 1) / c->slices) & cmask;
103 dest = dst + sstart * stride;
106 for (j = sstart; j < send; j++) {
107 for (i = 0; i < width * step; i += step) {
124 for (slice = 0; slice < c->slices; slice++) {
126 int slice_data_start, slice_data_end, slice_size;
129 send = (height * (slice + 1) / c->slices) & cmask;
130 dest = dst + sstart * stride;
132 // slice offset and size validation was done earlier
133 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
134 slice_data_end = AV_RL32(src + slice * 4);
135 slice_size = slice_data_end - slice_data_start;
138 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
139 "yet a slice has a length of zero.\n");
143 memcpy(c->slice_bits, src + slice_data_start + c->slices * 4,
145 memset(c->slice_bits + slice_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
146 c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
147 (uint32_t *) c->slice_bits,
148 (slice_data_end - slice_data_start + 3) >> 2);
149 init_get_bits(&gb, c->slice_bits, slice_size * 8);
152 for (j = sstart; j < send; j++) {
153 for (i = 0; i < width * step; i += step) {
154 if (get_bits_left(&gb) <= 0) {
155 av_log(c->avctx, AV_LOG_ERROR,
156 "Slice decoding ran out of bits\n");
159 pix = get_vlc2(&gb, vlc.table, vlc.bits, 3);
161 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
172 if (get_bits_left(&gb) > 32)
173 av_log(c->avctx, AV_LOG_WARNING,
174 "%d bits left after decoding slice\n", get_bits_left(&gb));
182 return AVERROR_INVALIDDATA;
185 static void restore_rgb_planes(uint8_t *src, int step, int stride, int width,
191 for (j = 0; j < height; j++) {
192 for (i = 0; i < width * step; i += step) {
196 src[i] = r + g - 0x80;
197 src[i + 2] = b + g - 0x80;
203 static void restore_median(uint8_t *src, int step, int stride,
204 int width, int height, int slices, int rmode)
209 int slice_start, slice_height;
210 const int cmask = ~rmode;
212 for (slice = 0; slice < slices; slice++) {
213 slice_start = ((slice * height) / slices) & cmask;
214 slice_height = ((((slice + 1) * height) / slices) & cmask) -
217 bsrc = src + slice_start * stride;
219 // first line - left neighbour prediction
222 for (i = step; i < width * step; i += step) {
227 if (slice_height == 1)
229 // second line - first element has top prediction, the rest uses median
233 for (i = step; i < width * step; i += step) {
234 B = bsrc[i - stride];
235 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
240 // the rest of lines use continuous median prediction
241 for (j = 2; j < slice_height; j++) {
242 for (i = 0; i < width * step; i += step) {
243 B = bsrc[i - stride];
244 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
253 /* UtVideo interlaced mode treats every two lines as a single one,
254 * so restoring function should take care of possible padding between
255 * two parts of the same "line".
257 static void restore_median_il(uint8_t *src, int step, int stride,
258 int width, int height, int slices, int rmode)
263 int slice_start, slice_height;
264 const int cmask = ~(rmode ? 3 : 1);
265 const int stride2 = stride << 1;
267 for (slice = 0; slice < slices; slice++) {
268 slice_start = ((slice * height) / slices) & cmask;
269 slice_height = ((((slice + 1) * height) / slices) & cmask) -
273 bsrc = src + slice_start * stride;
275 // first line - left neighbour prediction
278 for (i = step; i < width * step; i += step) {
282 for (i = 0; i < width * step; i += step) {
283 bsrc[stride + i] += A;
284 A = bsrc[stride + i];
287 if (slice_height == 1)
289 // second line - first element has top prediction, the rest uses median
293 for (i = step; i < width * step; i += step) {
294 B = bsrc[i - stride2];
295 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
299 for (i = 0; i < width * step; i += step) {
300 B = bsrc[i - stride];
301 bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C));
303 A = bsrc[stride + i];
306 // the rest of lines use continuous median prediction
307 for (j = 2; j < slice_height; j++) {
308 for (i = 0; i < width * step; i += step) {
309 B = bsrc[i - stride2];
310 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
314 for (i = 0; i < width * step; i += step) {
315 B = bsrc[i - stride];
316 bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C));
318 A = bsrc[i + stride];
325 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
328 const uint8_t *buf = avpkt->data;
329 int buf_size = avpkt->size;
330 UtvideoContext *c = avctx->priv_data;
332 const uint8_t *plane_start[5];
333 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
336 ThreadFrame frame = { .f = data };
338 if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0)
341 /* parse plane structure to get frame flags and validate slice offsets */
342 bytestream2_init(&gb, buf, buf_size);
343 for (i = 0; i < c->planes; i++) {
344 plane_start[i] = gb.buffer;
345 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
346 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
347 return AVERROR_INVALIDDATA;
349 bytestream2_skipu(&gb, 256);
352 for (j = 0; j < c->slices; j++) {
353 slice_end = bytestream2_get_le32u(&gb);
354 slice_size = slice_end - slice_start;
355 if (slice_end < 0 || slice_size < 0 ||
356 bytestream2_get_bytes_left(&gb) < slice_end) {
357 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
358 return AVERROR_INVALIDDATA;
360 slice_start = slice_end;
361 max_slice_size = FFMAX(max_slice_size, slice_size);
363 plane_size = slice_end;
364 bytestream2_skipu(&gb, plane_size);
366 plane_start[c->planes] = gb.buffer;
367 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
368 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
369 return AVERROR_INVALIDDATA;
371 c->frame_info = bytestream2_get_le32u(&gb);
372 av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
375 c->frame_pred = (c->frame_info >> 8) & 3;
377 if (c->frame_pred == PRED_GRADIENT) {
378 avpriv_request_sample(avctx, "Frame with gradient prediction");
379 return AVERROR_PATCHWELCOME;
382 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
383 max_slice_size + FF_INPUT_BUFFER_PADDING_SIZE);
385 if (!c->slice_bits) {
386 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
387 return AVERROR(ENOMEM);
390 switch (c->avctx->pix_fmt) {
391 case AV_PIX_FMT_RGB24:
392 case AV_PIX_FMT_RGBA:
393 for (i = 0; i < c->planes; i++) {
394 ret = decode_plane(c, i, frame.f->data[0] + ff_ut_rgb_order[i],
395 c->planes, frame.f->linesize[0], avctx->width,
396 avctx->height, plane_start[i],
397 c->frame_pred == PRED_LEFT);
400 if (c->frame_pred == PRED_MEDIAN) {
401 if (!c->interlaced) {
402 restore_median(frame.f->data[0] + ff_ut_rgb_order[i],
403 c->planes, frame.f->linesize[0], avctx->width,
404 avctx->height, c->slices, 0);
406 restore_median_il(frame.f->data[0] + ff_ut_rgb_order[i],
407 c->planes, frame.f->linesize[0],
408 avctx->width, avctx->height, c->slices,
413 restore_rgb_planes(frame.f->data[0], c->planes, frame.f->linesize[0],
414 avctx->width, avctx->height);
416 case AV_PIX_FMT_YUV420P:
417 for (i = 0; i < 3; i++) {
418 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
419 avctx->width >> !!i, avctx->height >> !!i,
420 plane_start[i], c->frame_pred == PRED_LEFT);
423 if (c->frame_pred == PRED_MEDIAN) {
424 if (!c->interlaced) {
425 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
426 avctx->width >> !!i, avctx->height >> !!i,
429 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
431 avctx->height >> !!i,
437 case AV_PIX_FMT_YUV422P:
438 for (i = 0; i < 3; i++) {
439 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
440 avctx->width >> !!i, avctx->height,
441 plane_start[i], c->frame_pred == PRED_LEFT);
444 if (c->frame_pred == PRED_MEDIAN) {
445 if (!c->interlaced) {
446 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
447 avctx->width >> !!i, avctx->height,
450 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
451 avctx->width >> !!i, avctx->height,
459 frame.f->key_frame = 1;
460 frame.f->pict_type = AV_PICTURE_TYPE_I;
461 frame.f->interlaced_frame = !!c->interlaced;
465 /* always report that the buffer was completely consumed */
469 static av_cold int decode_init(AVCodecContext *avctx)
471 UtvideoContext * const c = avctx->priv_data;
475 ff_bswapdsp_init(&c->bdsp);
477 if (avctx->extradata_size < 16) {
478 av_log(avctx, AV_LOG_ERROR,
479 "Insufficient extradata size %d, should be at least 16\n",
480 avctx->extradata_size);
481 return AVERROR_INVALIDDATA;
484 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
485 avctx->extradata[3], avctx->extradata[2],
486 avctx->extradata[1], avctx->extradata[0]);
487 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
488 AV_RB32(avctx->extradata + 4));
489 c->frame_info_size = AV_RL32(avctx->extradata + 8);
490 c->flags = AV_RL32(avctx->extradata + 12);
492 if (c->frame_info_size != 4)
493 avpriv_request_sample(avctx, "Frame info not 4 bytes");
494 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
495 c->slices = (c->flags >> 24) + 1;
496 c->compression = c->flags & 1;
497 c->interlaced = c->flags & 0x800;
499 c->slice_bits_size = 0;
501 switch (avctx->codec_tag) {
502 case MKTAG('U', 'L', 'R', 'G'):
504 avctx->pix_fmt = AV_PIX_FMT_RGB24;
506 case MKTAG('U', 'L', 'R', 'A'):
508 avctx->pix_fmt = AV_PIX_FMT_RGBA;
510 case MKTAG('U', 'L', 'Y', '0'):
512 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
513 avctx->colorspace = AVCOL_SPC_BT470BG;
515 case MKTAG('U', 'L', 'Y', '2'):
517 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
518 avctx->colorspace = AVCOL_SPC_BT470BG;
520 case MKTAG('U', 'L', 'H', '0'):
522 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
523 avctx->colorspace = AVCOL_SPC_BT709;
525 case MKTAG('U', 'L', 'H', '2'):
527 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
528 avctx->colorspace = AVCOL_SPC_BT709;
531 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
533 return AVERROR_INVALIDDATA;
539 static av_cold int decode_end(AVCodecContext *avctx)
541 UtvideoContext * const c = avctx->priv_data;
543 av_freep(&c->slice_bits);
548 AVCodec ff_utvideo_decoder = {
550 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
551 .type = AVMEDIA_TYPE_VIDEO,
552 .id = AV_CODEC_ID_UTVIDEO,
553 .priv_data_size = sizeof(UtvideoContext),
556 .decode = decode_frame,
557 .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,