utvideodec: Fix single symbol mode decoding
[libav.git] / libavcodec / utvideodec.c
1 /*
2 * Ut Video decoder
3 * Copyright (c) 2011 Konstantin Shishkov
4 *
5 * This file is part of Libav.
6 *
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.
11 *
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.
16 *
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
20 */
21
22 /**
23 * @file
24 * Ut Video decoder
25 */
26
27 #include <stdlib.h>
28
29 #include "libavutil/intreadwrite.h"
30 #include "avcodec.h"
31 #include "bytestream.h"
32 #include "get_bits.h"
33 #include "dsputil.h"
34 #include "thread.h"
35 #include "utvideo.h"
36
37 static int build_huff(const uint8_t *src, VLC *vlc, int *fsym)
38 {
39 int i;
40 HuffEntry he[256];
41 int last;
42 uint32_t codes[256];
43 uint8_t bits[256];
44 uint8_t syms[256];
45 uint32_t code;
46
47 *fsym = -1;
48 for (i = 0; i < 256; i++) {
49 he[i].sym = i;
50 he[i].len = *src++;
51 }
52 qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);
53
54 if (!he[0].len) {
55 *fsym = he[0].sym;
56 return 0;
57 }
58 if (he[0].len > 32)
59 return -1;
60
61 last = 255;
62 while (he[last].len == 255 && last)
63 last--;
64
65 code = 1;
66 for (i = last; i >= 0; i--) {
67 codes[i] = code >> (32 - he[i].len);
68 bits[i] = he[i].len;
69 syms[i] = he[i].sym;
70 code += 0x80000000u >> (he[i].len - 1);
71 }
72
73 return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 9), last + 1,
74 bits, sizeof(*bits), sizeof(*bits),
75 codes, sizeof(*codes), sizeof(*codes),
76 syms, sizeof(*syms), sizeof(*syms), 0);
77 }
78
79 static int decode_plane(UtvideoContext *c, int plane_no,
80 uint8_t *dst, int step, int stride,
81 int width, int height,
82 const uint8_t *src, int use_pred)
83 {
84 int i, j, slice, pix;
85 int sstart, send;
86 VLC vlc;
87 GetBitContext gb;
88 int prev, fsym;
89 const int cmask = ~(!plane_no && c->avctx->pix_fmt == PIX_FMT_YUV420P);
90
91 if (build_huff(src, &vlc, &fsym)) {
92 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
93 return AVERROR_INVALIDDATA;
94 }
95 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
96 send = 0;
97 for (slice = 0; slice < c->slices; slice++) {
98 uint8_t *dest;
99
100 sstart = send;
101 send = (height * (slice + 1) / c->slices) & cmask;
102 dest = dst + sstart * stride;
103
104 prev = 0x80;
105 for (j = sstart; j < send; j++) {
106 for (i = 0; i < width * step; i += step) {
107 pix = fsym;
108 if (use_pred) {
109 prev += pix;
110 pix = prev;
111 }
112 dest[i] = pix;
113 }
114 dest += stride;
115 }
116 }
117 return 0;
118 }
119
120 src += 256;
121
122 send = 0;
123 for (slice = 0; slice < c->slices; slice++) {
124 uint8_t *dest;
125 int slice_data_start, slice_data_end, slice_size;
126
127 sstart = send;
128 send = (height * (slice + 1) / c->slices) & cmask;
129 dest = dst + sstart * stride;
130
131 // slice offset and size validation was done earlier
132 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
133 slice_data_end = AV_RL32(src + slice * 4);
134 slice_size = slice_data_end - slice_data_start;
135
136 if (!slice_size) {
137 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
138 "yet a slice has a length of zero.\n");
139 goto fail;
140 }
141
142 memcpy(c->slice_bits, src + slice_data_start + c->slices * 4,
143 slice_size);
144 memset(c->slice_bits + slice_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
145 c->dsp.bswap_buf((uint32_t *) c->slice_bits, (uint32_t *) c->slice_bits,
146 (slice_data_end - slice_data_start + 3) >> 2);
147 init_get_bits(&gb, c->slice_bits, slice_size * 8);
148
149 prev = 0x80;
150 for (j = sstart; j < send; j++) {
151 for (i = 0; i < width * step; i += step) {
152 if (get_bits_left(&gb) <= 0) {
153 av_log(c->avctx, AV_LOG_ERROR,
154 "Slice decoding ran out of bits\n");
155 goto fail;
156 }
157 pix = get_vlc2(&gb, vlc.table, vlc.bits, 4);
158 if (pix < 0) {
159 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
160 goto fail;
161 }
162 if (use_pred) {
163 prev += pix;
164 pix = prev;
165 }
166 dest[i] = pix;
167 }
168 dest += stride;
169 }
170 if (get_bits_left(&gb) > 32)
171 av_log(c->avctx, AV_LOG_WARNING,
172 "%d bits left after decoding slice\n", get_bits_left(&gb));
173 }
174
175 ff_free_vlc(&vlc);
176
177 return 0;
178 fail:
179 ff_free_vlc(&vlc);
180 return AVERROR_INVALIDDATA;
181 }
182
183 static void restore_rgb_planes(uint8_t *src, int step, int stride, int width,
184 int height)
185 {
186 int i, j;
187 uint8_t r, g, b;
188
189 for (j = 0; j < height; j++) {
190 for (i = 0; i < width * step; i += step) {
191 r = src[i];
192 g = src[i + 1];
193 b = src[i + 2];
194 src[i] = r + g - 0x80;
195 src[i + 2] = b + g - 0x80;
196 }
197 src += stride;
198 }
199 }
200
201 static void restore_median(uint8_t *src, int step, int stride,
202 int width, int height, int slices, int rmode)
203 {
204 int i, j, slice;
205 int A, B, C;
206 uint8_t *bsrc;
207 int slice_start, slice_height;
208 const int cmask = ~rmode;
209
210 for (slice = 0; slice < slices; slice++) {
211 slice_start = ((slice * height) / slices) & cmask;
212 slice_height = ((((slice + 1) * height) / slices) & cmask) -
213 slice_start;
214
215 bsrc = src + slice_start * stride;
216
217 // first line - left neighbour prediction
218 bsrc[0] += 0x80;
219 A = bsrc[0];
220 for (i = step; i < width * step; i += step) {
221 bsrc[i] += A;
222 A = bsrc[i];
223 }
224 bsrc += stride;
225 if (slice_height == 1)
226 continue;
227 // second line - first element has top prediction, the rest uses median
228 C = bsrc[-stride];
229 bsrc[0] += C;
230 A = bsrc[0];
231 for (i = step; i < width * step; i += step) {
232 B = bsrc[i - stride];
233 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
234 C = B;
235 A = bsrc[i];
236 }
237 bsrc += stride;
238 // the rest of lines use continuous median prediction
239 for (j = 2; j < slice_height; j++) {
240 for (i = 0; i < width * step; i += step) {
241 B = bsrc[i - stride];
242 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
243 C = B;
244 A = bsrc[i];
245 }
246 bsrc += stride;
247 }
248 }
249 }
250
251 /* UtVideo interlaced mode treats every two lines as a single one,
252 * so restoring function should take care of possible padding between
253 * two parts of the same "line".
254 */
255 static void restore_median_il(uint8_t *src, int step, int stride,
256 int width, int height, int slices, int rmode)
257 {
258 int i, j, slice;
259 int A, B, C;
260 uint8_t *bsrc;
261 int slice_start, slice_height;
262 const int cmask = ~(rmode ? 3 : 1);
263 const int stride2 = stride << 1;
264
265 for (slice = 0; slice < slices; slice++) {
266 slice_start = ((slice * height) / slices) & cmask;
267 slice_height = ((((slice + 1) * height) / slices) & cmask) -
268 slice_start;
269 slice_height >>= 1;
270
271 bsrc = src + slice_start * stride;
272
273 // first line - left neighbour prediction
274 bsrc[0] += 0x80;
275 A = bsrc[0];
276 for (i = step; i < width * step; i += step) {
277 bsrc[i] += A;
278 A = bsrc[i];
279 }
280 for (i = 0; i < width * step; i += step) {
281 bsrc[stride + i] += A;
282 A = bsrc[stride + i];
283 }
284 bsrc += stride2;
285 if (slice_height == 1)
286 continue;
287 // second line - first element has top prediction, the rest uses median
288 C = bsrc[-stride2];
289 bsrc[0] += C;
290 A = bsrc[0];
291 for (i = step; i < width * step; i += step) {
292 B = bsrc[i - stride2];
293 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
294 C = B;
295 A = bsrc[i];
296 }
297 for (i = 0; i < width * step; i += step) {
298 B = bsrc[i - stride];
299 bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C));
300 C = B;
301 A = bsrc[stride + i];
302 }
303 bsrc += stride2;
304 // the rest of lines use continuous median prediction
305 for (j = 2; j < slice_height; j++) {
306 for (i = 0; i < width * step; i += step) {
307 B = bsrc[i - stride2];
308 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
309 C = B;
310 A = bsrc[i];
311 }
312 for (i = 0; i < width * step; i += step) {
313 B = bsrc[i - stride];
314 bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C));
315 C = B;
316 A = bsrc[i + stride];
317 }
318 bsrc += stride2;
319 }
320 }
321 }
322
323 static int decode_frame(AVCodecContext *avctx, void *data, int *data_size,
324 AVPacket *avpkt)
325 {
326 const uint8_t *buf = avpkt->data;
327 int buf_size = avpkt->size;
328 UtvideoContext *c = avctx->priv_data;
329 int i, j;
330 const uint8_t *plane_start[5];
331 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
332 int ret;
333 GetByteContext gb;
334
335 if (c->pic.data[0])
336 ff_thread_release_buffer(avctx, &c->pic);
337
338 c->pic.reference = 1;
339 c->pic.buffer_hints = FF_BUFFER_HINTS_VALID;
340 if ((ret = ff_thread_get_buffer(avctx, &c->pic)) < 0) {
341 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
342 return ret;
343 }
344
345 ff_thread_finish_setup(avctx);
346
347 /* parse plane structure to get frame flags and validate slice offsets */
348 bytestream2_init(&gb, buf, buf_size);
349 for (i = 0; i < c->planes; i++) {
350 plane_start[i] = gb.buffer;
351 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
352 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
353 return AVERROR_INVALIDDATA;
354 }
355 bytestream2_skipu(&gb, 256);
356 slice_start = 0;
357 slice_end = 0;
358 for (j = 0; j < c->slices; j++) {
359 slice_end = bytestream2_get_le32u(&gb);
360 slice_size = slice_end - slice_start;
361 if (slice_end < 0 || slice_size < 0 ||
362 bytestream2_get_bytes_left(&gb) < slice_end) {
363 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
364 return AVERROR_INVALIDDATA;
365 }
366 slice_start = slice_end;
367 max_slice_size = FFMAX(max_slice_size, slice_size);
368 }
369 plane_size = slice_end;
370 bytestream2_skipu(&gb, plane_size);
371 }
372 plane_start[c->planes] = gb.buffer;
373 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
374 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
375 return AVERROR_INVALIDDATA;
376 }
377 c->frame_info = bytestream2_get_le32u(&gb);
378 av_log(avctx, AV_LOG_DEBUG, "frame information flags %X\n", c->frame_info);
379
380 c->frame_pred = (c->frame_info >> 8) & 3;
381
382 if (c->frame_pred == PRED_GRADIENT) {
383 av_log_ask_for_sample(avctx, "Frame uses gradient prediction\n");
384 return AVERROR_PATCHWELCOME;
385 }
386
387 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
388 max_slice_size + FF_INPUT_BUFFER_PADDING_SIZE);
389
390 if (!c->slice_bits) {
391 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
392 return AVERROR(ENOMEM);
393 }
394
395 switch (c->avctx->pix_fmt) {
396 case PIX_FMT_RGB24:
397 case PIX_FMT_RGBA:
398 for (i = 0; i < c->planes; i++) {
399 ret = decode_plane(c, i, c->pic.data[0] + ff_ut_rgb_order[i],
400 c->planes, c->pic.linesize[0], avctx->width,
401 avctx->height, plane_start[i],
402 c->frame_pred == PRED_LEFT);
403 if (ret)
404 return ret;
405 if (c->frame_pred == PRED_MEDIAN) {
406 if (!c->interlaced) {
407 restore_median(c->pic.data[0] + ff_ut_rgb_order[i],
408 c->planes, c->pic.linesize[0], avctx->width,
409 avctx->height, c->slices, 0);
410 } else {
411 restore_median_il(c->pic.data[0] + ff_ut_rgb_order[i],
412 c->planes, c->pic.linesize[0],
413 avctx->width, avctx->height, c->slices,
414 0);
415 }
416 }
417 }
418 restore_rgb_planes(c->pic.data[0], c->planes, c->pic.linesize[0],
419 avctx->width, avctx->height);
420 break;
421 case PIX_FMT_YUV420P:
422 for (i = 0; i < 3; i++) {
423 ret = decode_plane(c, i, c->pic.data[i], 1, c->pic.linesize[i],
424 avctx->width >> !!i, avctx->height >> !!i,
425 plane_start[i], c->frame_pred == PRED_LEFT);
426 if (ret)
427 return ret;
428 if (c->frame_pred == PRED_MEDIAN) {
429 if (!c->interlaced) {
430 restore_median(c->pic.data[i], 1, c->pic.linesize[i],
431 avctx->width >> !!i, avctx->height >> !!i,
432 c->slices, !i);
433 } else {
434 restore_median_il(c->pic.data[i], 1, c->pic.linesize[i],
435 avctx->width >> !!i,
436 avctx->height >> !!i,
437 c->slices, !i);
438 }
439 }
440 }
441 break;
442 case PIX_FMT_YUV422P:
443 for (i = 0; i < 3; i++) {
444 ret = decode_plane(c, i, c->pic.data[i], 1, c->pic.linesize[i],
445 avctx->width >> !!i, avctx->height,
446 plane_start[i], c->frame_pred == PRED_LEFT);
447 if (ret)
448 return ret;
449 if (c->frame_pred == PRED_MEDIAN) {
450 if (!c->interlaced) {
451 restore_median(c->pic.data[i], 1, c->pic.linesize[i],
452 avctx->width >> !!i, avctx->height,
453 c->slices, 0);
454 } else {
455 restore_median_il(c->pic.data[i], 1, c->pic.linesize[i],
456 avctx->width >> !!i, avctx->height,
457 c->slices, 0);
458 }
459 }
460 }
461 break;
462 }
463
464 c->pic.key_frame = 1;
465 c->pic.pict_type = AV_PICTURE_TYPE_I;
466 c->pic.interlaced_frame = !!c->interlaced;
467
468 *data_size = sizeof(AVFrame);
469 *(AVFrame*)data = c->pic;
470
471 /* always report that the buffer was completely consumed */
472 return buf_size;
473 }
474
475 static av_cold int decode_init(AVCodecContext *avctx)
476 {
477 UtvideoContext * const c = avctx->priv_data;
478
479 c->avctx = avctx;
480
481 ff_dsputil_init(&c->dsp, avctx);
482
483 if (avctx->extradata_size < 16) {
484 av_log(avctx, AV_LOG_ERROR,
485 "Insufficient extradata size %d, should be at least 16\n",
486 avctx->extradata_size);
487 return AVERROR_INVALIDDATA;
488 }
489
490 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
491 avctx->extradata[3], avctx->extradata[2],
492 avctx->extradata[1], avctx->extradata[0]);
493 av_log(avctx, AV_LOG_DEBUG, "Original format %X\n",
494 AV_RB32(avctx->extradata + 4));
495 c->frame_info_size = AV_RL32(avctx->extradata + 8);
496 c->flags = AV_RL32(avctx->extradata + 12);
497
498 if (c->frame_info_size != 4)
499 av_log_ask_for_sample(avctx, "Frame info is not 4 bytes\n");
500 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08X\n", c->flags);
501 c->slices = (c->flags >> 24) + 1;
502 c->compression = c->flags & 1;
503 c->interlaced = c->flags & 0x800;
504
505 c->slice_bits_size = 0;
506
507 switch (avctx->codec_tag) {
508 case MKTAG('U', 'L', 'R', 'G'):
509 c->planes = 3;
510 avctx->pix_fmt = PIX_FMT_RGB24;
511 break;
512 case MKTAG('U', 'L', 'R', 'A'):
513 c->planes = 4;
514 avctx->pix_fmt = PIX_FMT_RGBA;
515 break;
516 case MKTAG('U', 'L', 'Y', '0'):
517 c->planes = 3;
518 avctx->pix_fmt = PIX_FMT_YUV420P;
519 break;
520 case MKTAG('U', 'L', 'Y', '2'):
521 c->planes = 3;
522 avctx->pix_fmt = PIX_FMT_YUV422P;
523 break;
524 default:
525 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
526 avctx->codec_tag);
527 return AVERROR_INVALIDDATA;
528 }
529
530 return 0;
531 }
532
533 static av_cold int decode_end(AVCodecContext *avctx)
534 {
535 UtvideoContext * const c = avctx->priv_data;
536
537 if (c->pic.data[0])
538 ff_thread_release_buffer(avctx, &c->pic);
539
540 av_freep(&c->slice_bits);
541
542 return 0;
543 }
544
545 AVCodec ff_utvideo_decoder = {
546 .name = "utvideo",
547 .type = AVMEDIA_TYPE_VIDEO,
548 .id = AV_CODEC_ID_UTVIDEO,
549 .priv_data_size = sizeof(UtvideoContext),
550 .init = decode_init,
551 .close = decode_end,
552 .decode = decode_frame,
553 .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
554 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
555 };