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"
32 #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->dsp.bswap_buf((uint32_t *) c->slice_bits, (uint32_t *) c->slice_bits,
147 (slice_data_end - slice_data_start + 3) >> 2);
148 init_get_bits(&gb, c->slice_bits, slice_size * 8);
151 for (j = sstart; j < send; j++) {
152 for (i = 0; i < width * step; i += step) {
153 if (get_bits_left(&gb) <= 0) {
154 av_log(c->avctx, AV_LOG_ERROR,
155 "Slice decoding ran out of bits\n");
158 pix = get_vlc2(&gb, vlc.table, vlc.bits, 3);
160 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
171 if (get_bits_left(&gb) > 32)
172 av_log(c->avctx, AV_LOG_WARNING,
173 "%d bits left after decoding slice\n", get_bits_left(&gb));
181 return AVERROR_INVALIDDATA;
184 static void restore_rgb_planes(uint8_t *src, int step, int stride, int width,
190 for (j = 0; j < height; j++) {
191 for (i = 0; i < width * step; i += step) {
195 src[i] = r + g - 0x80;
196 src[i + 2] = b + g - 0x80;
202 static void restore_median(uint8_t *src, int step, int stride,
203 int width, int height, int slices, int rmode)
208 int slice_start, slice_height;
209 const int cmask = ~rmode;
211 for (slice = 0; slice < slices; slice++) {
212 slice_start = ((slice * height) / slices) & cmask;
213 slice_height = ((((slice + 1) * height) / slices) & cmask) -
216 bsrc = src + slice_start * stride;
218 // first line - left neighbour prediction
221 for (i = step; i < width * step; i += step) {
226 if (slice_height == 1)
228 // second line - first element has top prediction, the rest uses median
232 for (i = step; i < width * step; i += step) {
233 B = bsrc[i - stride];
234 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
239 // the rest of lines use continuous median prediction
240 for (j = 2; j < slice_height; j++) {
241 for (i = 0; i < width * step; i += step) {
242 B = bsrc[i - stride];
243 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
252 /* UtVideo interlaced mode treats every two lines as a single one,
253 * so restoring function should take care of possible padding between
254 * two parts of the same "line".
256 static void restore_median_il(uint8_t *src, int step, int stride,
257 int width, int height, int slices, int rmode)
262 int slice_start, slice_height;
263 const int cmask = ~(rmode ? 3 : 1);
264 const int stride2 = stride << 1;
266 for (slice = 0; slice < slices; slice++) {
267 slice_start = ((slice * height) / slices) & cmask;
268 slice_height = ((((slice + 1) * height) / slices) & cmask) -
272 bsrc = src + slice_start * stride;
274 // first line - left neighbour prediction
277 for (i = step; i < width * step; i += step) {
281 for (i = 0; i < width * step; i += step) {
282 bsrc[stride + i] += A;
283 A = bsrc[stride + i];
286 if (slice_height == 1)
288 // second line - first element has top prediction, the rest uses median
292 for (i = step; i < width * step; i += step) {
293 B = bsrc[i - stride2];
294 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
298 for (i = 0; i < width * step; i += step) {
299 B = bsrc[i - stride];
300 bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C));
302 A = bsrc[stride + i];
305 // the rest of lines use continuous median prediction
306 for (j = 2; j < slice_height; j++) {
307 for (i = 0; i < width * step; i += step) {
308 B = bsrc[i - stride2];
309 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
313 for (i = 0; i < width * step; i += step) {
314 B = bsrc[i - stride];
315 bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C));
317 A = bsrc[i + stride];
324 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
327 const uint8_t *buf = avpkt->data;
328 int buf_size = avpkt->size;
329 UtvideoContext *c = avctx->priv_data;
331 const uint8_t *plane_start[5];
332 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
335 ThreadFrame frame = { .f = data };
337 if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0)
340 /* parse plane structure to get frame flags and validate slice offsets */
341 bytestream2_init(&gb, buf, buf_size);
342 for (i = 0; i < c->planes; i++) {
343 plane_start[i] = gb.buffer;
344 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
345 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
346 return AVERROR_INVALIDDATA;
348 bytestream2_skipu(&gb, 256);
351 for (j = 0; j < c->slices; j++) {
352 slice_end = bytestream2_get_le32u(&gb);
353 slice_size = slice_end - slice_start;
354 if (slice_end < 0 || slice_size < 0 ||
355 bytestream2_get_bytes_left(&gb) < slice_end) {
356 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
357 return AVERROR_INVALIDDATA;
359 slice_start = slice_end;
360 max_slice_size = FFMAX(max_slice_size, slice_size);
362 plane_size = slice_end;
363 bytestream2_skipu(&gb, plane_size);
365 plane_start[c->planes] = gb.buffer;
366 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
367 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
368 return AVERROR_INVALIDDATA;
370 c->frame_info = bytestream2_get_le32u(&gb);
371 av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
374 c->frame_pred = (c->frame_info >> 8) & 3;
376 if (c->frame_pred == PRED_GRADIENT) {
377 avpriv_request_sample(avctx, "Frame with gradient prediction");
378 return AVERROR_PATCHWELCOME;
381 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
382 max_slice_size + FF_INPUT_BUFFER_PADDING_SIZE);
384 if (!c->slice_bits) {
385 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
386 return AVERROR(ENOMEM);
389 switch (c->avctx->pix_fmt) {
390 case AV_PIX_FMT_RGB24:
391 case AV_PIX_FMT_RGBA:
392 for (i = 0; i < c->planes; i++) {
393 ret = decode_plane(c, i, frame.f->data[0] + ff_ut_rgb_order[i],
394 c->planes, frame.f->linesize[0], avctx->width,
395 avctx->height, plane_start[i],
396 c->frame_pred == PRED_LEFT);
399 if (c->frame_pred == PRED_MEDIAN) {
400 if (!c->interlaced) {
401 restore_median(frame.f->data[0] + ff_ut_rgb_order[i],
402 c->planes, frame.f->linesize[0], avctx->width,
403 avctx->height, c->slices, 0);
405 restore_median_il(frame.f->data[0] + ff_ut_rgb_order[i],
406 c->planes, frame.f->linesize[0],
407 avctx->width, avctx->height, c->slices,
412 restore_rgb_planes(frame.f->data[0], c->planes, frame.f->linesize[0],
413 avctx->width, avctx->height);
415 case AV_PIX_FMT_YUV420P:
416 for (i = 0; i < 3; i++) {
417 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
418 avctx->width >> !!i, avctx->height >> !!i,
419 plane_start[i], c->frame_pred == PRED_LEFT);
422 if (c->frame_pred == PRED_MEDIAN) {
423 if (!c->interlaced) {
424 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
425 avctx->width >> !!i, avctx->height >> !!i,
428 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
430 avctx->height >> !!i,
436 case AV_PIX_FMT_YUV422P:
437 for (i = 0; i < 3; i++) {
438 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
439 avctx->width >> !!i, avctx->height,
440 plane_start[i], c->frame_pred == PRED_LEFT);
443 if (c->frame_pred == PRED_MEDIAN) {
444 if (!c->interlaced) {
445 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
446 avctx->width >> !!i, avctx->height,
449 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
450 avctx->width >> !!i, avctx->height,
458 frame.f->key_frame = 1;
459 frame.f->pict_type = AV_PICTURE_TYPE_I;
460 frame.f->interlaced_frame = !!c->interlaced;
464 /* always report that the buffer was completely consumed */
468 static av_cold int decode_init(AVCodecContext *avctx)
470 UtvideoContext * const c = avctx->priv_data;
474 ff_dsputil_init(&c->dsp, avctx);
476 if (avctx->extradata_size < 16) {
477 av_log(avctx, AV_LOG_ERROR,
478 "Insufficient extradata size %d, should be at least 16\n",
479 avctx->extradata_size);
480 return AVERROR_INVALIDDATA;
483 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
484 avctx->extradata[3], avctx->extradata[2],
485 avctx->extradata[1], avctx->extradata[0]);
486 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
487 AV_RB32(avctx->extradata + 4));
488 c->frame_info_size = AV_RL32(avctx->extradata + 8);
489 c->flags = AV_RL32(avctx->extradata + 12);
491 if (c->frame_info_size != 4)
492 avpriv_request_sample(avctx, "Frame info not 4 bytes");
493 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
494 c->slices = (c->flags >> 24) + 1;
495 c->compression = c->flags & 1;
496 c->interlaced = c->flags & 0x800;
498 c->slice_bits_size = 0;
500 switch (avctx->codec_tag) {
501 case MKTAG('U', 'L', 'R', 'G'):
503 avctx->pix_fmt = AV_PIX_FMT_RGB24;
505 case MKTAG('U', 'L', 'R', 'A'):
507 avctx->pix_fmt = AV_PIX_FMT_RGBA;
509 case MKTAG('U', 'L', 'Y', '0'):
511 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
512 avctx->colorspace = AVCOL_SPC_BT470BG;
514 case MKTAG('U', 'L', 'Y', '2'):
516 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
517 avctx->colorspace = AVCOL_SPC_BT470BG;
519 case MKTAG('U', 'L', 'H', '0'):
521 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
522 avctx->colorspace = AVCOL_SPC_BT709;
524 case MKTAG('U', 'L', 'H', '2'):
526 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
527 avctx->colorspace = AVCOL_SPC_BT709;
530 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
532 return AVERROR_INVALIDDATA;
538 static av_cold int decode_end(AVCodecContext *avctx)
540 UtvideoContext * const c = avctx->priv_data;
542 av_freep(&c->slice_bits);
547 AVCodec ff_utvideo_decoder = {
549 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
550 .type = AVMEDIA_TYPE_VIDEO,
551 .id = AV_CODEC_ID_UTVIDEO,
552 .priv_data_size = sizeof(UtvideoContext),
555 .decode = decode_frame,
556 .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,