2 * Indeo Video Interactive v4 compatible decoder
3 * Copyright (c) 2009-2011 Maxim Poliakovski
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
24 * Indeo Video Interactive version 4 decoder
26 * Indeo 4 data is usually transported within .avi or .mov files.
27 * Known FOURCCs: 'IV41'
30 #define BITSTREAM_READER_LE
34 #include "ivi_common.h"
35 #include "indeo4data.h"
38 * Indeo 4 frame types.
42 FRAMETYPE_INTRA1 = 1, ///< intra frame with slightly different bitstream coding
43 FRAMETYPE_INTER = 2, ///< non-droppable P-frame
44 FRAMETYPE_BIDIR = 3, ///< bidirectional frame
45 FRAMETYPE_INTER_NOREF = 4, ///< droppable P-frame
46 FRAMETYPE_NULL_FIRST = 5, ///< empty frame with no data
47 FRAMETYPE_NULL_LAST = 6 ///< empty frame with no data
50 #define IVI4_PIC_SIZE_ESC 7
54 InvTransformPtr *inv_trans;
55 DCTransformPtr *dc_trans;
58 { ff_ivi_inverse_haar_8x8, ff_ivi_dc_haar_2d, 1 },
59 { ff_ivi_row_haar8, ff_ivi_dc_haar_2d, 0 },
60 { ff_ivi_col_haar8, ff_ivi_dc_haar_2d, 0 },
61 { ff_ivi_put_pixels_8x8, ff_ivi_put_dc_pixel_8x8, 1 },
62 { ff_ivi_inverse_slant_8x8, ff_ivi_dc_slant_2d, 1 },
63 { ff_ivi_row_slant8, ff_ivi_dc_row_slant, 1 },
64 { ff_ivi_col_slant8, ff_ivi_dc_col_slant, 1 },
65 { NULL, NULL, 0 }, /* inverse DCT 8x8 */
66 { NULL, NULL, 0 }, /* inverse DCT 8x1 */
67 { NULL, NULL, 0 }, /* inverse DCT 1x8 */
68 { ff_ivi_inverse_haar_4x4, ff_ivi_dc_haar_2d, 1 },
69 { ff_ivi_inverse_slant_4x4, ff_ivi_dc_slant_2d, 1 },
70 { NULL, NULL, 0 }, /* no transform 4x4 */
71 { ff_ivi_row_haar4, ff_ivi_dc_haar_2d, 0 },
72 { ff_ivi_col_haar4, ff_ivi_dc_haar_2d, 0 },
73 { ff_ivi_row_slant4, ff_ivi_dc_row_slant, 0 },
74 { ff_ivi_col_slant4, ff_ivi_dc_col_slant, 0 },
75 { NULL, NULL, 0 }, /* inverse DCT 4x4 */
79 * Decode subdivision of a plane.
80 * This is a simplified version that checks for two supported subdivisions:
81 * - 1 wavelet band per plane, size factor 1:1, code pattern: 3
82 * - 4 wavelet bands per plane, size factor 1:4, code pattern: 2,3,3,3,3
83 * Anything else is either unsupported or corrupt.
85 * @param[in,out] gb the GetBit context
86 * @return number of wavelet bands or 0 on error
88 static int decode_plane_subdivision(GetBitContext *gb)
92 switch (get_bits(gb, 2)) {
96 for (i = 0; i < 4; i++)
97 if (get_bits(gb, 2) != 3)
105 static inline int scale_tile_size(int def_size, int size_factor)
107 return size_factor == 15 ? def_size : (size_factor + 1) << 5;
111 * Decode Indeo 4 picture header.
113 * @param[in,out] ctx pointer to the decoder context
114 * @param[in] avctx pointer to the AVCodecContext
115 * @return result code: 0 = OK, negative number = error
117 static int decode_pic_hdr(IVI45DecContext *ctx, AVCodecContext *avctx)
119 int pic_size_indx, i, p;
120 IVIPicConfig pic_conf;
122 if (get_bits(&ctx->gb, 18) != 0x3FFF8) {
123 av_log(avctx, AV_LOG_ERROR, "Invalid picture start code!\n");
124 return AVERROR_INVALIDDATA;
127 ctx->prev_frame_type = ctx->frame_type;
128 ctx->frame_type = get_bits(&ctx->gb, 3);
129 if (ctx->frame_type == 7) {
130 av_log(avctx, AV_LOG_ERROR, "Invalid frame type: %d\n", ctx->frame_type);
131 return AVERROR_INVALIDDATA;
134 #if IVI4_STREAM_ANALYSER
135 if (ctx->frame_type == FRAMETYPE_BIDIR)
136 ctx->has_b_frames = 1;
139 ctx->transp_status = get_bits1(&ctx->gb);
140 #if IVI4_STREAM_ANALYSER
141 if (ctx->transp_status) {
146 /* unknown bit: Mac decoder ignores this bit, XANIM returns error */
147 if (get_bits1(&ctx->gb)) {
148 av_log(avctx, AV_LOG_ERROR, "Sync bit is set!\n");
149 return AVERROR_INVALIDDATA;
152 ctx->data_size = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 24) : 0;
154 /* null frames don't contain anything else so we just return */
155 if (ctx->frame_type >= FRAMETYPE_NULL_FIRST) {
156 av_dlog(avctx, "Null frame encountered!\n");
160 /* Check key lock status. If enabled - ignore lock word. */
161 /* Usually we have to prompt the user for the password, but */
162 /* we don't do that because Indeo 4 videos can be decoded anyway */
163 if (get_bits1(&ctx->gb)) {
164 skip_bits_long(&ctx->gb, 32);
165 av_dlog(avctx, "Password-protected clip!\n");
168 pic_size_indx = get_bits(&ctx->gb, 3);
169 if (pic_size_indx == IVI4_PIC_SIZE_ESC) {
170 pic_conf.pic_height = get_bits(&ctx->gb, 16);
171 pic_conf.pic_width = get_bits(&ctx->gb, 16);
173 pic_conf.pic_height = ivi4_common_pic_sizes[pic_size_indx * 2 + 1];
174 pic_conf.pic_width = ivi4_common_pic_sizes[pic_size_indx * 2 ];
177 /* Decode tile dimensions. */
178 if (get_bits1(&ctx->gb)) {
179 pic_conf.tile_height = scale_tile_size(pic_conf.pic_height, get_bits(&ctx->gb, 4));
180 pic_conf.tile_width = scale_tile_size(pic_conf.pic_width, get_bits(&ctx->gb, 4));
181 #if IVI4_STREAM_ANALYSER
182 ctx->uses_tiling = 1;
185 pic_conf.tile_height = pic_conf.pic_height;
186 pic_conf.tile_width = pic_conf.pic_width;
189 /* Decode chroma subsampling. We support only 4:4 aka YVU9. */
190 if (get_bits(&ctx->gb, 2)) {
191 av_log(avctx, AV_LOG_ERROR, "Only YVU9 picture format is supported!\n");
192 return AVERROR_INVALIDDATA;
194 pic_conf.chroma_height = (pic_conf.pic_height + 3) >> 2;
195 pic_conf.chroma_width = (pic_conf.pic_width + 3) >> 2;
197 /* decode subdivision of the planes */
198 pic_conf.luma_bands = decode_plane_subdivision(&ctx->gb);
199 if (pic_conf.luma_bands)
200 pic_conf.chroma_bands = decode_plane_subdivision(&ctx->gb);
201 ctx->is_scalable = pic_conf.luma_bands != 1 || pic_conf.chroma_bands != 1;
202 if (ctx->is_scalable && (pic_conf.luma_bands != 4 || pic_conf.chroma_bands != 1)) {
203 av_log(avctx, AV_LOG_ERROR, "Scalability: unsupported subdivision! Luma bands: %d, chroma bands: %d\n",
204 pic_conf.luma_bands, pic_conf.chroma_bands);
205 return AVERROR_INVALIDDATA;
208 /* check if picture layout was changed and reallocate buffers */
209 if (ivi_pic_config_cmp(&pic_conf, &ctx->pic_conf)) {
210 if (ff_ivi_init_planes(ctx->planes, &pic_conf)) {
211 av_log(avctx, AV_LOG_ERROR, "Couldn't reallocate color planes!\n");
212 ctx->pic_conf.luma_bands = 0;
213 return AVERROR(ENOMEM);
216 ctx->pic_conf = pic_conf;
218 /* set default macroblock/block dimensions */
219 for (p = 0; p <= 2; p++) {
220 for (i = 0; i < (!p ? pic_conf.luma_bands : pic_conf.chroma_bands); i++) {
221 ctx->planes[p].bands[i].mb_size = !p ? (!ctx->is_scalable ? 16 : 8) : 4;
222 ctx->planes[p].bands[i].blk_size = !p ? 8 : 4;
226 if (ff_ivi_init_tiles(ctx->planes, ctx->pic_conf.tile_width,
227 ctx->pic_conf.tile_height)) {
228 av_log(avctx, AV_LOG_ERROR,
229 "Couldn't reallocate internal structures!\n");
230 return AVERROR(ENOMEM);
234 ctx->frame_num = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 20) : 0;
236 /* skip decTimeEst field if present */
237 if (get_bits1(&ctx->gb))
238 skip_bits(&ctx->gb, 8);
240 /* decode macroblock and block huffman codebooks */
241 if (ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_MB_HUFF, &ctx->mb_vlc, avctx) ||
242 ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_BLK_HUFF, &ctx->blk_vlc, avctx))
243 return AVERROR_INVALIDDATA;
245 ctx->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
247 ctx->in_imf = get_bits1(&ctx->gb);
248 ctx->in_q = get_bits1(&ctx->gb);
250 ctx->pic_glob_quant = get_bits(&ctx->gb, 5);
252 /* TODO: ignore this parameter if unused */
253 ctx->unknown1 = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 0;
255 ctx->checksum = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 16) : 0;
257 /* skip picture header extension if any */
258 while (get_bits1(&ctx->gb)) {
259 av_dlog(avctx, "Pic hdr extension encountered!\n");
260 skip_bits(&ctx->gb, 8);
263 if (get_bits1(&ctx->gb)) {
264 av_log(avctx, AV_LOG_ERROR, "Bad blocks bits encountered!\n");
267 align_get_bits(&ctx->gb);
274 * Decode Indeo 4 band header.
276 * @param[in,out] ctx pointer to the decoder context
277 * @param[in,out] band pointer to the band descriptor
278 * @param[in] avctx pointer to the AVCodecContext
279 * @return result code: 0 = OK, negative number = error
281 static int decode_band_hdr(IVI45DecContext *ctx, IVIBandDesc *band,
282 AVCodecContext *avctx)
284 int plane, band_num, indx, transform_id, scan_indx;
287 plane = get_bits(&ctx->gb, 2);
288 band_num = get_bits(&ctx->gb, 4);
289 if (band->plane != plane || band->band_num != band_num) {
290 av_log(avctx, AV_LOG_ERROR, "Invalid band header sequence!\n");
291 return AVERROR_INVALIDDATA;
294 band->is_empty = get_bits1(&ctx->gb);
295 if (!band->is_empty) {
297 * If header size is not given, header size is 4 bytes. */
298 if (get_bits1(&ctx->gb))
299 skip_bits(&ctx->gb, 16);
301 band->is_halfpel = get_bits(&ctx->gb, 2);
302 if (band->is_halfpel >= 2) {
303 av_log(avctx, AV_LOG_ERROR, "Invalid/unsupported mv resolution: %d!\n",
305 return AVERROR_INVALIDDATA;
307 #if IVI4_STREAM_ANALYSER
308 if (!band->is_halfpel)
309 ctx->uses_fullpel = 1;
312 band->checksum_present = get_bits1(&ctx->gb);
313 if (band->checksum_present)
314 band->checksum = get_bits(&ctx->gb, 16);
316 indx = get_bits(&ctx->gb, 2);
318 av_log(avctx, AV_LOG_ERROR, "Invalid block size!\n");
319 return AVERROR_INVALIDDATA;
321 band->mb_size = 16 >> indx;
322 band->blk_size = 8 >> (indx >> 1);
324 band->inherit_mv = get_bits1(&ctx->gb);
325 band->inherit_qdelta = get_bits1(&ctx->gb);
327 band->glob_quant = get_bits(&ctx->gb, 5);
329 if (!get_bits1(&ctx->gb) || ctx->frame_type == FRAMETYPE_INTRA) {
330 transform_id = get_bits(&ctx->gb, 5);
331 if (transform_id >= FF_ARRAY_ELEMS(transforms) ||
332 !transforms[transform_id].inv_trans) {
333 avpriv_request_sample(avctx, "Transform %d", transform_id);
334 return AVERROR_PATCHWELCOME;
336 if ((transform_id >= 7 && transform_id <= 9) ||
337 transform_id == 17) {
338 avpriv_request_sample(avctx, "DCT transform");
339 return AVERROR_PATCHWELCOME;
342 #if IVI4_STREAM_ANALYSER
343 if ((transform_id >= 0 && transform_id <= 2) || transform_id == 10)
347 band->inv_transform = transforms[transform_id].inv_trans;
348 band->dc_transform = transforms[transform_id].dc_trans;
349 band->is_2d_trans = transforms[transform_id].is_2d_trans;
350 if (transform_id < 10)
351 band->transform_size = 8;
353 band->transform_size = 4;
355 if (band->blk_size != band->transform_size)
356 return AVERROR_INVALIDDATA;
358 scan_indx = get_bits(&ctx->gb, 4);
359 if (scan_indx == 15) {
360 av_log(avctx, AV_LOG_ERROR, "Custom scan pattern encountered!\n");
361 return AVERROR_INVALIDDATA;
363 if (scan_indx > 4 && scan_indx < 10) {
364 if (band->blk_size != 4)
365 return AVERROR_INVALIDDATA;
366 } else if (band->blk_size != 8)
367 return AVERROR_INVALIDDATA;
369 band->scan = scan_index_to_tab[scan_indx];
371 band->quant_mat = get_bits(&ctx->gb, 5);
372 if (band->quant_mat == 31) {
373 av_log(avctx, AV_LOG_ERROR, "Custom quant matrix encountered!\n");
374 return AVERROR_INVALIDDATA;
376 if (band->quant_mat >= FF_ARRAY_ELEMS(quant_index_to_tab)) {
377 avpriv_request_sample(avctx, "Quantization matrix %d",
379 return AVERROR_INVALIDDATA;
383 /* decode block huffman codebook */
384 if (!get_bits1(&ctx->gb))
385 band->blk_vlc.tab = ctx->blk_vlc.tab;
387 if (ff_ivi_dec_huff_desc(&ctx->gb, 1, IVI_BLK_HUFF,
388 &band->blk_vlc, avctx))
389 return AVERROR_INVALIDDATA;
391 /* select appropriate rvmap table for this band */
392 band->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
394 /* decode rvmap probability corrections if any */
395 band->num_corr = 0; /* there is no corrections */
396 if (get_bits1(&ctx->gb)) {
397 band->num_corr = get_bits(&ctx->gb, 8); /* get number of correction pairs */
398 if (band->num_corr > 61) {
399 av_log(avctx, AV_LOG_ERROR, "Too many corrections: %d\n",
401 return AVERROR_INVALIDDATA;
404 /* read correction pairs */
405 for (i = 0; i < band->num_corr * 2; i++)
406 band->corr[i] = get_bits(&ctx->gb, 8);
410 if (band->blk_size == 8) {
411 band->intra_base = &ivi4_quant_8x8_intra[quant_index_to_tab[band->quant_mat]][0];
412 band->inter_base = &ivi4_quant_8x8_inter[quant_index_to_tab[band->quant_mat]][0];
414 band->intra_base = &ivi4_quant_4x4_intra[quant_index_to_tab[band->quant_mat]][0];
415 band->inter_base = &ivi4_quant_4x4_inter[quant_index_to_tab[band->quant_mat]][0];
418 /* Indeo 4 doesn't use scale tables */
419 band->intra_scale = NULL;
420 band->inter_scale = NULL;
422 align_get_bits(&ctx->gb);
429 * Decode information (block type, cbp, quant delta, motion vector)
430 * for all macroblocks in the current tile.
432 * @param[in,out] ctx pointer to the decoder context
433 * @param[in,out] band pointer to the band descriptor
434 * @param[in,out] tile pointer to the tile descriptor
435 * @param[in] avctx pointer to the AVCodecContext
436 * @return result code: 0 = OK, negative number = error
438 static int decode_mb_info(IVI45DecContext *ctx, IVIBandDesc *band,
439 IVITile *tile, AVCodecContext *avctx)
441 int x, y, mv_x, mv_y, mv_delta, offs, mb_offset, blks_per_mb,
442 mv_scale, mb_type_bits;
443 IVIMbInfo *mb, *ref_mb;
444 int row_offset = band->mb_size * band->pitch;
447 ref_mb = tile->ref_mbs;
448 offs = tile->ypos * band->pitch + tile->xpos;
450 blks_per_mb = band->mb_size != band->blk_size ? 4 : 1;
451 mb_type_bits = ctx->frame_type == FRAMETYPE_BIDIR ? 2 : 1;
453 /* scale factor for motion vectors */
454 mv_scale = (ctx->planes[0].bands[0].mb_size >> 3) - (band->mb_size >> 3);
457 for (y = tile->ypos; y < tile->ypos + tile->height; y += band->mb_size) {
460 for (x = tile->xpos; x < tile->xpos + tile->width; x += band->mb_size) {
463 mb->buf_offs = mb_offset;
465 if (get_bits1(&ctx->gb)) {
466 if (ctx->frame_type == FRAMETYPE_INTRA) {
467 av_log(avctx, AV_LOG_ERROR, "Empty macroblock in an INTRA picture!\n");
468 return AVERROR_INVALIDDATA;
470 mb->type = 1; /* empty macroblocks are always INTER */
471 mb->cbp = 0; /* all blocks are empty */
474 if (!band->plane && !band->band_num && ctx->in_q) {
475 mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
477 mb->q_delta = IVI_TOSIGNED(mb->q_delta);
480 mb->mv_x = mb->mv_y = 0; /* no motion vector coded */
481 if (band->inherit_mv && ref_mb) {
482 /* motion vector inheritance */
484 mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
485 mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
487 mb->mv_x = ref_mb->mv_x;
488 mb->mv_y = ref_mb->mv_y;
492 if (band->inherit_mv) {
493 /* copy mb_type from corresponding reference mb */
495 return AVERROR_INVALIDDATA;
496 mb->type = ref_mb->type;
497 } else if (ctx->frame_type == FRAMETYPE_INTRA ||
498 ctx->frame_type == FRAMETYPE_INTRA1) {
499 mb->type = 0; /* mb_type is always INTRA for intra-frames */
501 mb->type = get_bits(&ctx->gb, mb_type_bits);
504 mb->cbp = get_bits(&ctx->gb, blks_per_mb);
507 if (band->inherit_qdelta) {
508 if (ref_mb) mb->q_delta = ref_mb->q_delta;
509 } else if (mb->cbp || (!band->plane && !band->band_num &&
511 mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
513 mb->q_delta = IVI_TOSIGNED(mb->q_delta);
517 mb->mv_x = mb->mv_y = 0; /* there is no motion vector in intra-macroblocks */
519 if (band->inherit_mv) {
521 /* motion vector inheritance */
523 mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
524 mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
526 mb->mv_x = ref_mb->mv_x;
527 mb->mv_y = ref_mb->mv_y;
530 /* decode motion vector deltas */
531 mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
533 mv_y += IVI_TOSIGNED(mv_delta);
534 mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
536 mv_x += IVI_TOSIGNED(mv_delta);
546 mb_offset += band->mb_size;
552 align_get_bits(&ctx->gb);
559 * Rearrange decoding and reference buffers.
561 * @param[in,out] ctx pointer to the decoder context
563 static void switch_buffers(IVI45DecContext *ctx)
565 switch (ctx->prev_frame_type) {
566 case FRAMETYPE_INTRA:
567 case FRAMETYPE_INTRA1:
568 case FRAMETYPE_INTER:
569 ctx->buf_switch ^= 1;
570 ctx->dst_buf = ctx->buf_switch;
571 ctx->ref_buf = ctx->buf_switch ^ 1;
573 case FRAMETYPE_INTER_NOREF:
577 switch (ctx->frame_type) {
578 case FRAMETYPE_INTRA:
579 case FRAMETYPE_INTRA1:
582 case FRAMETYPE_INTER:
583 ctx->dst_buf = ctx->buf_switch;
584 ctx->ref_buf = ctx->buf_switch ^ 1;
586 case FRAMETYPE_INTER_NOREF:
587 case FRAMETYPE_NULL_FIRST:
588 case FRAMETYPE_NULL_LAST:
594 static int is_nonnull_frame(IVI45DecContext *ctx)
596 return ctx->frame_type < FRAMETYPE_NULL_FIRST;
600 static av_cold int decode_init(AVCodecContext *avctx)
602 IVI45DecContext *ctx = avctx->priv_data;
604 ff_ivi_init_static_vlc();
606 /* copy rvmap tables in our context so we can apply changes to them */
607 memcpy(ctx->rvmap_tabs, ff_ivi_rvmap_tabs, sizeof(ff_ivi_rvmap_tabs));
609 /* Force allocation of the internal buffers */
610 /* during picture header decoding. */
611 ctx->pic_conf.pic_width = 0;
612 ctx->pic_conf.pic_height = 0;
614 avctx->pix_fmt = AV_PIX_FMT_YUV410P;
616 ctx->decode_pic_hdr = decode_pic_hdr;
617 ctx->decode_band_hdr = decode_band_hdr;
618 ctx->decode_mb_info = decode_mb_info;
619 ctx->switch_buffers = switch_buffers;
620 ctx->is_nonnull_frame = is_nonnull_frame;
626 AVCodec ff_indeo4_decoder = {
628 .type = AVMEDIA_TYPE_VIDEO,
629 .id = AV_CODEC_ID_INDEO4,
630 .priv_data_size = sizeof(IVI45DecContext),
632 .close = ff_ivi_decode_close,
633 .decode = ff_ivi_decode_frame,
634 .long_name = NULL_IF_CONFIG_SMALL("Intel Indeo Video Interactive 4"),
635 .capabilities = CODEC_CAP_DR1,