e641a7583059da019bd22afa7cb0ae397ad7a820
[libav.git] / libavcodec / vqavideo.c
1 /*
2 * Westwood Studios VQA Video Decoder
3 * Copyright (C) 2003 the ffmpeg project
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 * VQA Video Decoder
25 * @author Mike Melanson (melanson@pcisys.net)
26 * @see http://wiki.multimedia.cx/index.php?title=VQA
27 *
28 * The VQA video decoder outputs PAL8 or RGB555 colorspace data, depending
29 * on the type of data in the file.
30 *
31 * This decoder needs the 42-byte VQHD header from the beginning
32 * of the VQA file passed through the extradata field. The VQHD header
33 * is laid out as:
34 *
35 * bytes 0-3 chunk fourcc: 'VQHD'
36 * bytes 4-7 chunk size in big-endian format, should be 0x0000002A
37 * bytes 8-49 VQHD chunk data
38 *
39 * Bytes 8-49 are what this decoder expects to see.
40 *
41 * Briefly, VQA is a vector quantized animation format that operates in a
42 * VGA palettized colorspace. It operates on pixel vectors (blocks)
43 * of either 4x2 or 4x4 in size. Compressed VQA chunks can contain vector
44 * codebooks, palette information, and code maps for rendering vectors onto
45 * frames. Any of these components can also be compressed with a run-length
46 * encoding (RLE) algorithm commonly referred to as "format80".
47 *
48 * VQA takes a novel approach to rate control. Each group of n frames
49 * (usually, n = 8) relies on a different vector codebook. Rather than
50 * transporting an entire codebook every 8th frame, the new codebook is
51 * broken up into 8 pieces and sent along with the compressed video chunks
52 * for each of the 8 frames preceding the 8 frames which require the
53 * codebook. A full codebook is also sent on the very first frame of a
54 * file. This is an interesting technique, although it makes random file
55 * seeking difficult despite the fact that the frames are all intracoded.
56 *
57 * V1,2 VQA uses 12-bit codebook indexes. If the 12-bit indexes were
58 * packed into bytes and then RLE compressed, bytewise, the results would
59 * be poor. That is why the coding method divides each index into 2 parts,
60 * the top 4 bits and the bottom 8 bits, then RL encodes the 4-bit pieces
61 * together and the 8-bit pieces together. If most of the vectors are
62 * clustered into one group of 256 vectors, most of the 4-bit index pieces
63 * should be the same.
64 */
65
66 #include <stdio.h>
67 #include <stdlib.h>
68 #include <string.h>
69
70 #include "libavutil/intreadwrite.h"
71 #include "libavutil/imgutils.h"
72 #include "avcodec.h"
73 #include "bytestream.h"
74 #include "internal.h"
75
76 #define PALETTE_COUNT 256
77 #define VQA_HEADER_SIZE 0x2A
78
79 /* allocate the maximum vector space, regardless of the file version:
80 * (0xFF00 codebook vectors + 0x100 solid pixel vectors) * (4x4 pixels/block) */
81 #define MAX_CODEBOOK_VECTORS 0xFF00
82 #define SOLID_PIXEL_VECTORS 0x100
83 #define MAX_VECTORS (MAX_CODEBOOK_VECTORS + SOLID_PIXEL_VECTORS)
84 #define MAX_CODEBOOK_SIZE (MAX_VECTORS * 4 * 4)
85
86 #define CBF0_TAG MKBETAG('C', 'B', 'F', '0')
87 #define CBFZ_TAG MKBETAG('C', 'B', 'F', 'Z')
88 #define CBP0_TAG MKBETAG('C', 'B', 'P', '0')
89 #define CBPZ_TAG MKBETAG('C', 'B', 'P', 'Z')
90 #define CPL0_TAG MKBETAG('C', 'P', 'L', '0')
91 #define CPLZ_TAG MKBETAG('C', 'P', 'L', 'Z')
92 #define VPTZ_TAG MKBETAG('V', 'P', 'T', 'Z')
93
94 typedef struct VqaContext {
95
96 AVCodecContext *avctx;
97 AVFrame frame;
98 GetByteContext gb;
99
100 uint32_t palette[PALETTE_COUNT];
101
102 int width; /* width of a frame */
103 int height; /* height of a frame */
104 int vector_width; /* width of individual vector */
105 int vector_height; /* height of individual vector */
106 int vqa_version; /* this should be either 1, 2 or 3 */
107
108 unsigned char *codebook; /* the current codebook */
109 int codebook_size;
110 unsigned char *next_codebook_buffer; /* accumulator for next codebook */
111 int next_codebook_buffer_index;
112
113 unsigned char *decode_buffer;
114 int decode_buffer_size;
115
116 /* number of frames to go before replacing codebook */
117 int partial_countdown;
118 int partial_count;
119
120 } VqaContext;
121
122 static av_cold int vqa_decode_init(AVCodecContext *avctx)
123 {
124 VqaContext *s = avctx->priv_data;
125 int i, j, codebook_index;
126
127 s->avctx = avctx;
128 avctx->pix_fmt = AV_PIX_FMT_PAL8;
129
130 /* make sure the extradata made it */
131 if (s->avctx->extradata_size != VQA_HEADER_SIZE) {
132 av_log(s->avctx, AV_LOG_ERROR, " VQA video: expected extradata size of %d\n", VQA_HEADER_SIZE);
133 return -1;
134 }
135
136 /* load up the VQA parameters from the header */
137 s->vqa_version = s->avctx->extradata[0];
138 s->width = AV_RL16(&s->avctx->extradata[6]);
139 s->height = AV_RL16(&s->avctx->extradata[8]);
140 if(av_image_check_size(s->width, s->height, 0, avctx)){
141 s->width= s->height= 0;
142 return -1;
143 }
144 s->vector_width = s->avctx->extradata[10];
145 s->vector_height = s->avctx->extradata[11];
146 s->partial_count = s->partial_countdown = s->avctx->extradata[13];
147
148 /* the vector dimensions have to meet very stringent requirements */
149 if ((s->vector_width != 4) ||
150 ((s->vector_height != 2) && (s->vector_height != 4))) {
151 /* return without further initialization */
152 return -1;
153 }
154
155 if (s->width & (s->vector_width - 1) ||
156 s->height & (s->vector_height - 1)) {
157 av_log(avctx, AV_LOG_ERROR, "Image size not multiple of block size\n");
158 return AVERROR_INVALIDDATA;
159 }
160
161 /* allocate codebooks */
162 s->codebook_size = MAX_CODEBOOK_SIZE;
163 s->codebook = av_malloc(s->codebook_size);
164 if (!s->codebook)
165 goto fail;
166 s->next_codebook_buffer = av_malloc(s->codebook_size);
167 if (!s->next_codebook_buffer)
168 goto fail;
169
170 /* allocate decode buffer */
171 s->decode_buffer_size = (s->width / s->vector_width) *
172 (s->height / s->vector_height) * 2;
173 s->decode_buffer = av_malloc(s->decode_buffer_size);
174 if (!s->decode_buffer)
175 goto fail;
176
177 /* initialize the solid-color vectors */
178 if (s->vector_height == 4) {
179 codebook_index = 0xFF00 * 16;
180 for (i = 0; i < 256; i++)
181 for (j = 0; j < 16; j++)
182 s->codebook[codebook_index++] = i;
183 } else {
184 codebook_index = 0xF00 * 8;
185 for (i = 0; i < 256; i++)
186 for (j = 0; j < 8; j++)
187 s->codebook[codebook_index++] = i;
188 }
189 s->next_codebook_buffer_index = 0;
190
191 s->frame.data[0] = NULL;
192
193 return 0;
194 fail:
195 av_freep(&s->codebook);
196 av_freep(&s->next_codebook_buffer);
197 av_freep(&s->decode_buffer);
198 return AVERROR(ENOMEM);
199 }
200
201 #define CHECK_COUNT() \
202 if (dest_index + count > dest_size) { \
203 av_log(NULL, AV_LOG_ERROR, " VQA video: decode_format80 problem: next op would overflow dest_index\n"); \
204 av_log(NULL, AV_LOG_ERROR, " VQA video: current dest_index = %d, count = %d, dest_size = %d\n", \
205 dest_index, count, dest_size); \
206 return AVERROR_INVALIDDATA; \
207 }
208
209 #define CHECK_COPY(idx) \
210 if (idx < 0 || idx + count > dest_size) { \
211 av_log(NULL, AV_LOG_ERROR, " VQA video: decode_format80 problem: next op would overflow dest_index\n"); \
212 av_log(NULL, AV_LOG_ERROR, " VQA video: current src_pos = %d, count = %d, dest_size = %d\n", \
213 src_pos, count, dest_size); \
214 return AVERROR_INVALIDDATA; \
215 }
216
217
218 static int decode_format80(GetByteContext *gb, int src_size,
219 unsigned char *dest, int dest_size, int check_size) {
220
221 int dest_index = 0;
222 int count, opcode, start;
223 int src_pos;
224 unsigned char color;
225 int i;
226
227 start = bytestream2_tell(gb);
228 while (bytestream2_tell(gb) - start < src_size) {
229 opcode = bytestream2_get_byte(gb);
230 av_dlog(NULL, " opcode %02X: ", opcode);
231
232 /* 0x80 means that frame is finished */
233 if (opcode == 0x80)
234 return 0;
235
236 if (dest_index >= dest_size) {
237 av_log(NULL, AV_LOG_ERROR, " VQA video: decode_format80 problem: dest_index (%d) exceeded dest_size (%d)\n",
238 dest_index, dest_size);
239 return AVERROR_INVALIDDATA;
240 }
241
242 if (opcode == 0xFF) {
243
244 count = bytestream2_get_le16(gb);
245 src_pos = bytestream2_get_le16(gb);
246 av_dlog(NULL, "(1) copy %X bytes from absolute pos %X\n", count, src_pos);
247 CHECK_COUNT();
248 CHECK_COPY(src_pos);
249 for (i = 0; i < count; i++)
250 dest[dest_index + i] = dest[src_pos + i];
251 dest_index += count;
252
253 } else if (opcode == 0xFE) {
254
255 count = bytestream2_get_le16(gb);
256 color = bytestream2_get_byte(gb);
257 av_dlog(NULL, "(2) set %X bytes to %02X\n", count, color);
258 CHECK_COUNT();
259 memset(&dest[dest_index], color, count);
260 dest_index += count;
261
262 } else if ((opcode & 0xC0) == 0xC0) {
263
264 count = (opcode & 0x3F) + 3;
265 src_pos = bytestream2_get_le16(gb);
266 av_dlog(NULL, "(3) copy %X bytes from absolute pos %X\n", count, src_pos);
267 CHECK_COUNT();
268 CHECK_COPY(src_pos);
269 for (i = 0; i < count; i++)
270 dest[dest_index + i] = dest[src_pos + i];
271 dest_index += count;
272
273 } else if (opcode > 0x80) {
274
275 count = opcode & 0x3F;
276 av_dlog(NULL, "(4) copy %X bytes from source to dest\n", count);
277 CHECK_COUNT();
278 bytestream2_get_buffer(gb, &dest[dest_index], count);
279 dest_index += count;
280
281 } else {
282
283 count = ((opcode & 0x70) >> 4) + 3;
284 src_pos = bytestream2_get_byte(gb) | ((opcode & 0x0F) << 8);
285 av_dlog(NULL, "(5) copy %X bytes from relpos %X\n", count, src_pos);
286 CHECK_COUNT();
287 CHECK_COPY(dest_index - src_pos);
288 for (i = 0; i < count; i++)
289 dest[dest_index + i] = dest[dest_index - src_pos + i];
290 dest_index += count;
291 }
292 }
293
294 /* validate that the entire destination buffer was filled; this is
295 * important for decoding frame maps since each vector needs to have a
296 * codebook entry; it is not important for compressed codebooks because
297 * not every entry needs to be filled */
298 if (check_size)
299 if (dest_index < dest_size)
300 av_log(NULL, AV_LOG_ERROR, " VQA video: decode_format80 problem: decode finished with dest_index (%d) < dest_size (%d)\n",
301 dest_index, dest_size);
302
303 return 0; // let's display what we decoded anyway
304 }
305
306 static int vqa_decode_chunk(VqaContext *s)
307 {
308 unsigned int chunk_type;
309 unsigned int chunk_size;
310 int byte_skip;
311 unsigned int index = 0;
312 int i;
313 unsigned char r, g, b;
314 int index_shift;
315 int res;
316
317 int cbf0_chunk = -1;
318 int cbfz_chunk = -1;
319 int cbp0_chunk = -1;
320 int cbpz_chunk = -1;
321 int cpl0_chunk = -1;
322 int cplz_chunk = -1;
323 int vptz_chunk = -1;
324
325 int x, y;
326 int lines = 0;
327 int pixel_ptr;
328 int vector_index = 0;
329 int lobyte = 0;
330 int hibyte = 0;
331 int lobytes = 0;
332 int hibytes = s->decode_buffer_size / 2;
333
334 /* first, traverse through the frame and find the subchunks */
335 while (bytestream2_get_bytes_left(&s->gb) >= 8) {
336
337 chunk_type = bytestream2_get_be32u(&s->gb);
338 index = bytestream2_tell(&s->gb);
339 chunk_size = bytestream2_get_be32u(&s->gb);
340
341 switch (chunk_type) {
342
343 case CBF0_TAG:
344 cbf0_chunk = index;
345 break;
346
347 case CBFZ_TAG:
348 cbfz_chunk = index;
349 break;
350
351 case CBP0_TAG:
352 cbp0_chunk = index;
353 break;
354
355 case CBPZ_TAG:
356 cbpz_chunk = index;
357 break;
358
359 case CPL0_TAG:
360 cpl0_chunk = index;
361 break;
362
363 case CPLZ_TAG:
364 cplz_chunk = index;
365 break;
366
367 case VPTZ_TAG:
368 vptz_chunk = index;
369 break;
370
371 default:
372 av_log(s->avctx, AV_LOG_ERROR, " VQA video: Found unknown chunk type: %c%c%c%c (%08X)\n",
373 (chunk_type >> 24) & 0xFF,
374 (chunk_type >> 16) & 0xFF,
375 (chunk_type >> 8) & 0xFF,
376 (chunk_type >> 0) & 0xFF,
377 chunk_type);
378 break;
379 }
380
381 byte_skip = chunk_size & 0x01;
382 bytestream2_skip(&s->gb, chunk_size + byte_skip);
383 }
384
385 /* next, deal with the palette */
386 if ((cpl0_chunk != -1) && (cplz_chunk != -1)) {
387
388 /* a chunk should not have both chunk types */
389 av_log(s->avctx, AV_LOG_ERROR, " VQA video: problem: found both CPL0 and CPLZ chunks\n");
390 return AVERROR_INVALIDDATA;
391 }
392
393 /* decompress the palette chunk */
394 if (cplz_chunk != -1) {
395
396 /* yet to be handled */
397
398 }
399
400 /* convert the RGB palette into the machine's endian format */
401 if (cpl0_chunk != -1) {
402
403 bytestream2_seek(&s->gb, cpl0_chunk, SEEK_SET);
404 chunk_size = bytestream2_get_be32(&s->gb);
405 /* sanity check the palette size */
406 if (chunk_size / 3 > 256 || chunk_size > bytestream2_get_bytes_left(&s->gb)) {
407 av_log(s->avctx, AV_LOG_ERROR, " VQA video: problem: found a palette chunk with %d colors\n",
408 chunk_size / 3);
409 return AVERROR_INVALIDDATA;
410 }
411 for (i = 0; i < chunk_size / 3; i++) {
412 /* scale by 4 to transform 6-bit palette -> 8-bit */
413 r = bytestream2_get_byteu(&s->gb) * 4;
414 g = bytestream2_get_byteu(&s->gb) * 4;
415 b = bytestream2_get_byteu(&s->gb) * 4;
416 s->palette[i] = (r << 16) | (g << 8) | (b);
417 }
418 }
419
420 /* next, look for a full codebook */
421 if ((cbf0_chunk != -1) && (cbfz_chunk != -1)) {
422
423 /* a chunk should not have both chunk types */
424 av_log(s->avctx, AV_LOG_ERROR, " VQA video: problem: found both CBF0 and CBFZ chunks\n");
425 return AVERROR_INVALIDDATA;
426 }
427
428 /* decompress the full codebook chunk */
429 if (cbfz_chunk != -1) {
430
431 bytestream2_seek(&s->gb, cbfz_chunk, SEEK_SET);
432 chunk_size = bytestream2_get_be32(&s->gb);
433 if ((res = decode_format80(&s->gb, chunk_size, s->codebook,
434 s->codebook_size, 0)) < 0)
435 return res;
436 }
437
438 /* copy a full codebook */
439 if (cbf0_chunk != -1) {
440
441 bytestream2_seek(&s->gb, cbf0_chunk, SEEK_SET);
442 chunk_size = bytestream2_get_be32(&s->gb);
443 /* sanity check the full codebook size */
444 if (chunk_size > MAX_CODEBOOK_SIZE) {
445 av_log(s->avctx, AV_LOG_ERROR, " VQA video: problem: CBF0 chunk too large (0x%X bytes)\n",
446 chunk_size);
447 return AVERROR_INVALIDDATA;
448 }
449
450 bytestream2_get_buffer(&s->gb, s->codebook, chunk_size);
451 }
452
453 /* decode the frame */
454 if (vptz_chunk == -1) {
455
456 /* something is wrong if there is no VPTZ chunk */
457 av_log(s->avctx, AV_LOG_ERROR, " VQA video: problem: no VPTZ chunk found\n");
458 return AVERROR_INVALIDDATA;
459 }
460
461 bytestream2_seek(&s->gb, vptz_chunk, SEEK_SET);
462 chunk_size = bytestream2_get_be32(&s->gb);
463 if ((res = decode_format80(&s->gb, chunk_size,
464 s->decode_buffer, s->decode_buffer_size, 1)) < 0)
465 return res;
466
467 /* render the final PAL8 frame */
468 if (s->vector_height == 4)
469 index_shift = 4;
470 else
471 index_shift = 3;
472 for (y = 0; y < s->height; y += s->vector_height) {
473 for (x = 0; x < s->width; x += 4, lobytes++, hibytes++) {
474 pixel_ptr = y * s->frame.linesize[0] + x;
475
476 /* get the vector index, the method for which varies according to
477 * VQA file version */
478 switch (s->vqa_version) {
479
480 case 1:
481 lobyte = s->decode_buffer[lobytes * 2];
482 hibyte = s->decode_buffer[(lobytes * 2) + 1];
483 vector_index = ((hibyte << 8) | lobyte) >> 3;
484 vector_index <<= index_shift;
485 lines = s->vector_height;
486 /* uniform color fill - a quick hack */
487 if (hibyte == 0xFF) {
488 while (lines--) {
489 s->frame.data[0][pixel_ptr + 0] = 255 - lobyte;
490 s->frame.data[0][pixel_ptr + 1] = 255 - lobyte;
491 s->frame.data[0][pixel_ptr + 2] = 255 - lobyte;
492 s->frame.data[0][pixel_ptr + 3] = 255 - lobyte;
493 pixel_ptr += s->frame.linesize[0];
494 }
495 lines=0;
496 }
497 break;
498
499 case 2:
500 lobyte = s->decode_buffer[lobytes];
501 hibyte = s->decode_buffer[hibytes];
502 vector_index = (hibyte << 8) | lobyte;
503 vector_index <<= index_shift;
504 lines = s->vector_height;
505 break;
506
507 case 3:
508 /* not implemented yet */
509 lines = 0;
510 break;
511 }
512
513 while (lines--) {
514 s->frame.data[0][pixel_ptr + 0] = s->codebook[vector_index++];
515 s->frame.data[0][pixel_ptr + 1] = s->codebook[vector_index++];
516 s->frame.data[0][pixel_ptr + 2] = s->codebook[vector_index++];
517 s->frame.data[0][pixel_ptr + 3] = s->codebook[vector_index++];
518 pixel_ptr += s->frame.linesize[0];
519 }
520 }
521 }
522
523 /* handle partial codebook */
524 if ((cbp0_chunk != -1) && (cbpz_chunk != -1)) {
525 /* a chunk should not have both chunk types */
526 av_log(s->avctx, AV_LOG_ERROR, " VQA video: problem: found both CBP0 and CBPZ chunks\n");
527 return AVERROR_INVALIDDATA;
528 }
529
530 if (cbp0_chunk != -1) {
531
532 bytestream2_seek(&s->gb, cbp0_chunk, SEEK_SET);
533 chunk_size = bytestream2_get_be32(&s->gb);
534
535 /* accumulate partial codebook */
536 bytestream2_get_buffer(&s->gb, &s->next_codebook_buffer[s->next_codebook_buffer_index],
537 chunk_size);
538 s->next_codebook_buffer_index += chunk_size;
539
540 s->partial_countdown--;
541 if (s->partial_countdown == 0) {
542
543 /* time to replace codebook */
544 memcpy(s->codebook, s->next_codebook_buffer,
545 s->next_codebook_buffer_index);
546
547 /* reset accounting */
548 s->next_codebook_buffer_index = 0;
549 s->partial_countdown = s->partial_count;
550 }
551 }
552
553 if (cbpz_chunk != -1) {
554
555 bytestream2_seek(&s->gb, cbpz_chunk, SEEK_SET);
556 chunk_size = bytestream2_get_be32(&s->gb);
557
558 /* accumulate partial codebook */
559 bytestream2_get_buffer(&s->gb, &s->next_codebook_buffer[s->next_codebook_buffer_index],
560 chunk_size);
561 s->next_codebook_buffer_index += chunk_size;
562
563 s->partial_countdown--;
564 if (s->partial_countdown == 0) {
565 GetByteContext gb;
566
567 bytestream2_init(&gb, s->next_codebook_buffer, s->next_codebook_buffer_index);
568 /* decompress codebook */
569 if ((res = decode_format80(&gb, s->next_codebook_buffer_index,
570 s->codebook, s->codebook_size, 0)) < 0)
571 return res;
572
573 /* reset accounting */
574 s->next_codebook_buffer_index = 0;
575 s->partial_countdown = s->partial_count;
576 }
577 }
578
579 return 0;
580 }
581
582 static int vqa_decode_frame(AVCodecContext *avctx,
583 void *data, int *data_size,
584 AVPacket *avpkt)
585 {
586 VqaContext *s = avctx->priv_data;
587 int res;
588
589 if (s->frame.data[0])
590 avctx->release_buffer(avctx, &s->frame);
591
592 if (ff_get_buffer(avctx, &s->frame)) {
593 av_log(s->avctx, AV_LOG_ERROR, " VQA Video: get_buffer() failed\n");
594 return -1;
595 }
596
597 bytestream2_init(&s->gb, avpkt->data, avpkt->size);
598 if ((res = vqa_decode_chunk(s)) < 0)
599 return res;
600
601 /* make the palette available on the way out */
602 memcpy(s->frame.data[1], s->palette, PALETTE_COUNT * 4);
603 s->frame.palette_has_changed = 1;
604
605 *data_size = sizeof(AVFrame);
606 *(AVFrame*)data = s->frame;
607
608 /* report that the buffer was completely consumed */
609 return avpkt->size;
610 }
611
612 static av_cold int vqa_decode_end(AVCodecContext *avctx)
613 {
614 VqaContext *s = avctx->priv_data;
615
616 av_freep(&s->codebook);
617 av_freep(&s->next_codebook_buffer);
618 av_freep(&s->decode_buffer);
619
620 if (s->frame.data[0])
621 avctx->release_buffer(avctx, &s->frame);
622
623 return 0;
624 }
625
626 AVCodec ff_vqa_decoder = {
627 .name = "vqavideo",
628 .type = AVMEDIA_TYPE_VIDEO,
629 .id = AV_CODEC_ID_WS_VQA,
630 .priv_data_size = sizeof(VqaContext),
631 .init = vqa_decode_init,
632 .close = vqa_decode_end,
633 .decode = vqa_decode_frame,
634 .capabilities = CODEC_CAP_DR1,
635 .long_name = NULL_IF_CONFIG_SMALL("Westwood Studios VQA (Vector Quantized Animation) video"),
636 };