3 * Copyright (c) 2011 Konstantin Shishkov
5 * This file is part of Libav.
7 * Libav 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 * Libav 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 Libav; 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 "bitstream.h"
35 #include "bytestream.h"
39 static int build_huff(const uint8_t *src, VLC *vlc, int *fsym)
50 for (i = 0; i < 256; i++) {
54 qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);
64 while (he[last].len == 255 && last)
68 for (i = last; i >= 0; i--) {
69 codes[i] = code >> (32 - he[i].len);
72 code += 0x80000000u >> (he[i].len - 1);
75 return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 9), last + 1,
76 bits, sizeof(*bits), sizeof(*bits),
77 codes, sizeof(*codes), sizeof(*codes),
78 syms, sizeof(*syms), sizeof(*syms), 0);
81 static int decode_plane(UtvideoContext *c, int plane_no,
82 uint8_t *dst, int step, ptrdiff_t stride,
83 int width, int height,
84 const uint8_t *src, int use_pred)
91 const int cmask = ~(!plane_no && c->avctx->pix_fmt == AV_PIX_FMT_YUV420P);
93 if (build_huff(src, &vlc, &fsym)) {
94 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
95 return AVERROR_INVALIDDATA;
97 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
99 for (slice = 0; slice < c->slices; slice++) {
103 send = (height * (slice + 1) / c->slices) & cmask;
104 dest = dst + sstart * stride;
107 for (j = sstart; j < send; j++) {
108 for (i = 0; i < width * step; i += step) {
125 for (slice = 0; slice < c->slices; slice++) {
127 int slice_data_start, slice_data_end, slice_size;
130 send = (height * (slice + 1) / c->slices) & cmask;
131 dest = dst + sstart * stride;
133 // slice offset and size validation was done earlier
134 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
135 slice_data_end = AV_RL32(src + slice * 4);
136 slice_size = slice_data_end - slice_data_start;
139 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
140 "yet a slice has a length of zero.\n");
144 memcpy(c->slice_bits, src + slice_data_start + c->slices * 4,
146 memset(c->slice_bits + slice_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
147 c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
148 (uint32_t *) c->slice_bits,
149 (slice_data_end - slice_data_start + 3) >> 2);
150 bitstream_init(&bc, c->slice_bits, slice_size * 8);
153 for (j = sstart; j < send; j++) {
154 for (i = 0; i < width * step; i += step) {
155 if (bitstream_bits_left(&bc) <= 0) {
156 av_log(c->avctx, AV_LOG_ERROR,
157 "Slice decoding ran out of bits\n");
160 pix = bitstream_read_vlc(&bc, vlc.table, vlc.bits, 4);
162 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
173 if (bitstream_bits_left(&bc) > 32)
174 av_log(c->avctx, AV_LOG_WARNING,
175 "%d bits left after decoding slice\n", bitstream_bits_left(&bc));
183 return AVERROR_INVALIDDATA;
186 static void restore_rgb_planes(uint8_t *src, int step, ptrdiff_t stride,
187 int width, int height)
192 for (j = 0; j < height; j++) {
193 for (i = 0; i < width * step; i += step) {
197 src[i] = r + g - 0x80;
198 src[i + 2] = b + g - 0x80;
204 static void restore_median(uint8_t *src, int step, ptrdiff_t stride,
205 int width, int height, int slices, int rmode)
210 int slice_start, slice_height;
211 const int cmask = ~rmode;
213 for (slice = 0; slice < slices; slice++) {
214 slice_start = ((slice * height) / slices) & cmask;
215 slice_height = ((((slice + 1) * height) / slices) & cmask) -
220 bsrc = src + slice_start * stride;
222 // first line - left neighbour prediction
225 for (i = step; i < width * step; i += step) {
230 if (slice_height == 1)
232 // second line - first element has top prediction, the rest uses median
236 for (i = step; i < width * step; i += step) {
237 B = bsrc[i - stride];
238 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
243 // the rest of lines use continuous median prediction
244 for (j = 2; j < slice_height; j++) {
245 for (i = 0; i < width * step; i += step) {
246 B = bsrc[i - stride];
247 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
256 /* UtVideo interlaced mode treats every two lines as a single one,
257 * so restoring function should take care of possible padding between
258 * two parts of the same "line".
260 static void restore_median_il(uint8_t *src, int step, ptrdiff_t stride,
261 int width, int height, int slices, int rmode)
266 int slice_start, slice_height;
267 const int cmask = ~(rmode ? 3 : 1);
268 const ptrdiff_t stride2 = stride << 1;
270 for (slice = 0; slice < slices; slice++) {
271 slice_start = ((slice * height) / slices) & cmask;
272 slice_height = ((((slice + 1) * height) / slices) & cmask) -
278 bsrc = src + slice_start * stride;
280 // first line - left neighbour prediction
283 for (i = step; i < width * step; i += step) {
287 for (i = 0; i < width * step; i += step) {
288 bsrc[stride + i] += A;
289 A = bsrc[stride + i];
292 if (slice_height == 1)
294 // second line - first element has top prediction, the rest uses median
298 for (i = step; i < width * step; i += step) {
299 B = bsrc[i - stride2];
300 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
304 for (i = 0; i < width * step; i += step) {
305 B = bsrc[i - stride];
306 bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C));
308 A = bsrc[stride + i];
311 // the rest of lines use continuous median prediction
312 for (j = 2; j < slice_height; j++) {
313 for (i = 0; i < width * step; i += step) {
314 B = bsrc[i - stride2];
315 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
319 for (i = 0; i < width * step; i += step) {
320 B = bsrc[i - stride];
321 bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C));
323 A = bsrc[i + stride];
330 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
333 const uint8_t *buf = avpkt->data;
334 int buf_size = avpkt->size;
335 UtvideoContext *c = avctx->priv_data;
337 const uint8_t *plane_start[5];
338 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
341 ThreadFrame frame = { .f = data };
343 if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) {
344 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
348 ff_thread_finish_setup(avctx);
350 /* parse plane structure to get frame flags and validate slice offsets */
351 bytestream2_init(&gb, buf, buf_size);
352 for (i = 0; i < c->planes; i++) {
353 plane_start[i] = gb.buffer;
354 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
355 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
356 return AVERROR_INVALIDDATA;
358 bytestream2_skipu(&gb, 256);
361 for (j = 0; j < c->slices; j++) {
362 slice_end = bytestream2_get_le32u(&gb);
363 slice_size = slice_end - slice_start;
364 if (slice_end < 0 || slice_size < 0 ||
365 bytestream2_get_bytes_left(&gb) < slice_end) {
366 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
367 return AVERROR_INVALIDDATA;
369 slice_start = slice_end;
370 max_slice_size = FFMAX(max_slice_size, slice_size);
372 plane_size = slice_end;
373 bytestream2_skipu(&gb, plane_size);
375 plane_start[c->planes] = gb.buffer;
376 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
377 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
378 return AVERROR_INVALIDDATA;
380 c->frame_info = bytestream2_get_le32u(&gb);
381 av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
384 c->frame_pred = (c->frame_info >> 8) & 3;
386 if (c->frame_pred == PRED_GRADIENT) {
387 avpriv_request_sample(avctx, "Frame with gradient prediction");
388 return AVERROR_PATCHWELCOME;
391 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
392 max_slice_size + AV_INPUT_BUFFER_PADDING_SIZE);
394 if (!c->slice_bits) {
395 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
396 return AVERROR(ENOMEM);
399 switch (c->avctx->pix_fmt) {
400 case AV_PIX_FMT_RGB24:
401 case AV_PIX_FMT_RGBA:
402 for (i = 0; i < c->planes; i++) {
403 ret = decode_plane(c, i, frame.f->data[0] + ff_ut_rgb_order[i],
404 c->planes, frame.f->linesize[0], avctx->width,
405 avctx->height, plane_start[i],
406 c->frame_pred == PRED_LEFT);
409 if (c->frame_pred == PRED_MEDIAN) {
410 if (!c->interlaced) {
411 restore_median(frame.f->data[0] + ff_ut_rgb_order[i],
412 c->planes, frame.f->linesize[0], avctx->width,
413 avctx->height, c->slices, 0);
415 restore_median_il(frame.f->data[0] + ff_ut_rgb_order[i],
416 c->planes, frame.f->linesize[0],
417 avctx->width, avctx->height, c->slices,
422 restore_rgb_planes(frame.f->data[0], c->planes, frame.f->linesize[0],
423 avctx->width, avctx->height);
425 case AV_PIX_FMT_YUV420P:
426 for (i = 0; i < 3; i++) {
427 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
428 avctx->width >> !!i, avctx->height >> !!i,
429 plane_start[i], c->frame_pred == PRED_LEFT);
432 if (c->frame_pred == PRED_MEDIAN) {
433 if (!c->interlaced) {
434 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
435 avctx->width >> !!i, avctx->height >> !!i,
438 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
440 avctx->height >> !!i,
446 case AV_PIX_FMT_YUV422P:
447 for (i = 0; i < 3; i++) {
448 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
449 avctx->width >> !!i, avctx->height,
450 plane_start[i], c->frame_pred == PRED_LEFT);
453 if (c->frame_pred == PRED_MEDIAN) {
454 if (!c->interlaced) {
455 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
456 avctx->width >> !!i, avctx->height,
459 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
460 avctx->width >> !!i, avctx->height,
468 frame.f->key_frame = 1;
469 frame.f->pict_type = AV_PICTURE_TYPE_I;
470 frame.f->interlaced_frame = !!c->interlaced;
474 /* always report that the buffer was completely consumed */
478 static av_cold int decode_init(AVCodecContext *avctx)
480 UtvideoContext * const c = avctx->priv_data;
484 ff_bswapdsp_init(&c->bdsp);
486 if (avctx->extradata_size < 16) {
487 av_log(avctx, AV_LOG_ERROR,
488 "Insufficient extradata size %d, should be at least 16\n",
489 avctx->extradata_size);
490 return AVERROR_INVALIDDATA;
493 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
494 avctx->extradata[3], avctx->extradata[2],
495 avctx->extradata[1], avctx->extradata[0]);
496 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
497 AV_RB32(avctx->extradata + 4));
498 c->frame_info_size = AV_RL32(avctx->extradata + 8);
499 c->flags = AV_RL32(avctx->extradata + 12);
501 if (c->frame_info_size != 4)
502 avpriv_request_sample(avctx, "Frame info not 4 bytes");
503 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
504 c->slices = (c->flags >> 24) + 1;
505 c->compression = c->flags & 1;
506 c->interlaced = c->flags & 0x800;
508 c->slice_bits_size = 0;
510 switch (avctx->codec_tag) {
511 case MKTAG('U', 'L', 'R', 'G'):
513 avctx->pix_fmt = AV_PIX_FMT_RGB24;
515 case MKTAG('U', 'L', 'R', 'A'):
517 avctx->pix_fmt = AV_PIX_FMT_RGBA;
519 case MKTAG('U', 'L', 'Y', '0'):
521 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
522 avctx->colorspace = AVCOL_SPC_BT470BG;
524 case MKTAG('U', 'L', 'Y', '2'):
526 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
527 avctx->colorspace = AVCOL_SPC_BT470BG;
529 case MKTAG('U', 'L', 'H', '0'):
531 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
532 avctx->colorspace = AVCOL_SPC_BT709;
534 case MKTAG('U', 'L', 'H', '2'):
536 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
537 avctx->colorspace = AVCOL_SPC_BT709;
540 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
542 return AVERROR_INVALIDDATA;
548 static av_cold int decode_end(AVCodecContext *avctx)
550 UtvideoContext * const c = avctx->priv_data;
552 av_freep(&c->slice_bits);
557 AVCodec ff_utvideo_decoder = {
559 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
560 .type = AVMEDIA_TYPE_VIDEO,
561 .id = AV_CODEC_ID_UTVIDEO,
562 .priv_data_size = sizeof(UtvideoContext),
565 .decode = decode_frame,
566 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,