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
35 #include "indeo4data.h"
38 #define IVI4_PIC_SIZE_ESC 7
42 InvTransformPtr *inv_trans;
43 DCTransformPtr *dc_trans;
46 { ff_ivi_inverse_haar_8x8, ff_ivi_dc_haar_2d, 1 },
47 { ff_ivi_row_haar8, ff_ivi_dc_haar_2d, 0 },
48 { ff_ivi_col_haar8, ff_ivi_dc_haar_2d, 0 },
49 { ff_ivi_put_pixels_8x8, ff_ivi_put_dc_pixel_8x8, 1 },
50 { ff_ivi_inverse_slant_8x8, ff_ivi_dc_slant_2d, 1 },
51 { ff_ivi_row_slant8, ff_ivi_dc_row_slant, 1 },
52 { ff_ivi_col_slant8, ff_ivi_dc_col_slant, 1 },
53 { NULL, NULL, 0 }, /* inverse DCT 8x8 */
54 { NULL, NULL, 0 }, /* inverse DCT 8x1 */
55 { NULL, NULL, 0 }, /* inverse DCT 1x8 */
56 { ff_ivi_inverse_haar_4x4, ff_ivi_dc_haar_2d, 1 },
57 { ff_ivi_inverse_slant_4x4, ff_ivi_dc_slant_2d, 1 },
58 { NULL, NULL, 0 }, /* no transform 4x4 */
59 { ff_ivi_row_haar4, ff_ivi_dc_haar_2d, 0 },
60 { ff_ivi_col_haar4, ff_ivi_dc_haar_2d, 0 },
61 { ff_ivi_row_slant4, ff_ivi_dc_row_slant, 0 },
62 { ff_ivi_col_slant4, ff_ivi_dc_col_slant, 0 },
63 { NULL, NULL, 0 }, /* inverse DCT 4x4 */
67 * Decode subdivision of a plane.
68 * This is a simplified version that checks for two supported subdivisions:
69 * - 1 wavelet band per plane, size factor 1:1, code pattern: 3
70 * - 4 wavelet bands per plane, size factor 1:4, code pattern: 2,3,3,3,3
71 * Anything else is either unsupported or corrupt.
73 * @param[in,out] gb the GetBit context
74 * @return number of wavelet bands or 0 on error
76 static int decode_plane_subdivision(GetBitContext *gb)
80 switch (get_bits(gb, 2)) {
84 for (i = 0; i < 4; i++)
85 if (get_bits(gb, 2) != 3)
93 static inline int scale_tile_size(int def_size, int size_factor)
95 return size_factor == 15 ? def_size : (size_factor + 1) << 5;
99 * Decode Indeo 4 picture header.
101 * @param[in,out] ctx pointer to the decoder context
102 * @param[in] avctx pointer to the AVCodecContext
103 * @return result code: 0 = OK, negative number = error
105 static int decode_pic_hdr(IVI45DecContext *ctx, AVCodecContext *avctx)
107 int pic_size_indx, i, p;
108 IVIPicConfig pic_conf;
110 if (get_bits(&ctx->gb, 18) != 0x3FFF8) {
111 av_log(avctx, AV_LOG_ERROR, "Invalid picture start code!\n");
112 return AVERROR_INVALIDDATA;
115 ctx->prev_frame_type = ctx->frame_type;
116 ctx->frame_type = get_bits(&ctx->gb, 3);
117 if (ctx->frame_type == 7) {
118 av_log(avctx, AV_LOG_ERROR, "Invalid frame type: %d\n", ctx->frame_type);
119 return AVERROR_INVALIDDATA;
122 #if IVI4_STREAM_ANALYSER
123 if (ctx->frame_type == IVI4_FRAMETYPE_BIDIR)
124 ctx->has_b_frames = 1;
127 ctx->transp_status = get_bits1(&ctx->gb);
128 #if IVI4_STREAM_ANALYSER
129 if (ctx->transp_status) {
134 /* unknown bit: Mac decoder ignores this bit, XANIM returns error */
135 if (get_bits1(&ctx->gb)) {
136 av_log(avctx, AV_LOG_ERROR, "Sync bit is set!\n");
137 return AVERROR_INVALIDDATA;
140 ctx->data_size = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 24) : 0;
142 /* null frames don't contain anything else so we just return */
143 if (ctx->frame_type >= IVI4_FRAMETYPE_NULL_FIRST) {
144 ff_dlog(avctx, "Null frame encountered!\n");
148 /* Check key lock status. If enabled - ignore lock word. */
149 /* Usually we have to prompt the user for the password, but */
150 /* we don't do that because Indeo 4 videos can be decoded anyway */
151 if (get_bits1(&ctx->gb)) {
152 skip_bits_long(&ctx->gb, 32);
153 ff_dlog(avctx, "Password-protected clip!\n");
156 pic_size_indx = get_bits(&ctx->gb, 3);
157 if (pic_size_indx == IVI4_PIC_SIZE_ESC) {
158 pic_conf.pic_height = get_bits(&ctx->gb, 16);
159 pic_conf.pic_width = get_bits(&ctx->gb, 16);
161 pic_conf.pic_height = ivi4_common_pic_sizes[pic_size_indx * 2 + 1];
162 pic_conf.pic_width = ivi4_common_pic_sizes[pic_size_indx * 2 ];
165 /* Decode tile dimensions. */
166 if (get_bits1(&ctx->gb)) {
167 pic_conf.tile_height = scale_tile_size(pic_conf.pic_height, get_bits(&ctx->gb, 4));
168 pic_conf.tile_width = scale_tile_size(pic_conf.pic_width, get_bits(&ctx->gb, 4));
169 #if IVI4_STREAM_ANALYSER
170 ctx->uses_tiling = 1;
173 pic_conf.tile_height = pic_conf.pic_height;
174 pic_conf.tile_width = pic_conf.pic_width;
177 /* Decode chroma subsampling. We support only 4:4 aka YVU9. */
178 if (get_bits(&ctx->gb, 2)) {
179 av_log(avctx, AV_LOG_ERROR, "Only YVU9 picture format is supported!\n");
180 return AVERROR_INVALIDDATA;
182 pic_conf.chroma_height = (pic_conf.pic_height + 3) >> 2;
183 pic_conf.chroma_width = (pic_conf.pic_width + 3) >> 2;
185 /* decode subdivision of the planes */
186 pic_conf.luma_bands = decode_plane_subdivision(&ctx->gb);
187 if (pic_conf.luma_bands)
188 pic_conf.chroma_bands = decode_plane_subdivision(&ctx->gb);
189 ctx->is_scalable = pic_conf.luma_bands != 1 || pic_conf.chroma_bands != 1;
190 if (ctx->is_scalable && (pic_conf.luma_bands != 4 || pic_conf.chroma_bands != 1)) {
191 av_log(avctx, AV_LOG_ERROR, "Scalability: unsupported subdivision! Luma bands: %d, chroma bands: %d\n",
192 pic_conf.luma_bands, pic_conf.chroma_bands);
193 return AVERROR_INVALIDDATA;
196 /* check if picture layout was changed and reallocate buffers */
197 if (ivi_pic_config_cmp(&pic_conf, &ctx->pic_conf)) {
198 if (ff_ivi_init_planes(ctx->planes, &pic_conf, 1)) {
199 av_log(avctx, AV_LOG_ERROR, "Couldn't reallocate color planes!\n");
200 ctx->pic_conf.luma_bands = 0;
201 return AVERROR(ENOMEM);
204 ctx->pic_conf = pic_conf;
206 /* set default macroblock/block dimensions */
207 for (p = 0; p <= 2; p++) {
208 for (i = 0; i < (!p ? pic_conf.luma_bands : pic_conf.chroma_bands); i++) {
209 ctx->planes[p].bands[i].mb_size = !p ? (!ctx->is_scalable ? 16 : 8) : 4;
210 ctx->planes[p].bands[i].blk_size = !p ? 8 : 4;
214 if (ff_ivi_init_tiles(ctx->planes, ctx->pic_conf.tile_width,
215 ctx->pic_conf.tile_height)) {
216 av_log(avctx, AV_LOG_ERROR,
217 "Couldn't reallocate internal structures!\n");
218 return AVERROR(ENOMEM);
222 ctx->frame_num = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 20) : 0;
224 /* skip decTimeEst field if present */
225 if (get_bits1(&ctx->gb))
226 skip_bits(&ctx->gb, 8);
228 /* decode macroblock and block huffman codebooks */
229 if (ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_MB_HUFF, &ctx->mb_vlc, avctx) ||
230 ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_BLK_HUFF, &ctx->blk_vlc, avctx))
231 return AVERROR_INVALIDDATA;
233 ctx->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
235 ctx->in_imf = get_bits1(&ctx->gb);
236 ctx->in_q = get_bits1(&ctx->gb);
238 ctx->pic_glob_quant = get_bits(&ctx->gb, 5);
240 /* TODO: ignore this parameter if unused */
241 ctx->unknown1 = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 0;
243 ctx->checksum = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 16) : 0;
245 /* skip picture header extension if any */
246 while (get_bits1(&ctx->gb)) {
247 ff_dlog(avctx, "Pic hdr extension encountered!\n");
248 skip_bits(&ctx->gb, 8);
251 if (get_bits1(&ctx->gb)) {
252 av_log(avctx, AV_LOG_ERROR, "Bad blocks bits encountered!\n");
255 align_get_bits(&ctx->gb);
262 * Decode Indeo 4 band header.
264 * @param[in,out] ctx pointer to the decoder context
265 * @param[in,out] band pointer to the band descriptor
266 * @param[in] avctx pointer to the AVCodecContext
267 * @return result code: 0 = OK, negative number = error
269 static int decode_band_hdr(IVI45DecContext *ctx, IVIBandDesc *band,
270 AVCodecContext *avctx)
272 int plane, band_num, indx, transform_id, scan_indx;
275 plane = get_bits(&ctx->gb, 2);
276 band_num = get_bits(&ctx->gb, 4);
277 if (band->plane != plane || band->band_num != band_num) {
278 av_log(avctx, AV_LOG_ERROR, "Invalid band header sequence!\n");
279 return AVERROR_INVALIDDATA;
282 band->is_empty = get_bits1(&ctx->gb);
283 if (!band->is_empty) {
284 int old_blk_size = band->blk_size;
286 * If header size is not given, header size is 4 bytes. */
287 if (get_bits1(&ctx->gb))
288 skip_bits(&ctx->gb, 16);
290 band->is_halfpel = get_bits(&ctx->gb, 2);
291 if (band->is_halfpel >= 2) {
292 av_log(avctx, AV_LOG_ERROR, "Invalid/unsupported mv resolution: %d!\n",
294 return AVERROR_INVALIDDATA;
296 #if IVI4_STREAM_ANALYSER
297 if (!band->is_halfpel)
298 ctx->uses_fullpel = 1;
301 band->checksum_present = get_bits1(&ctx->gb);
302 if (band->checksum_present)
303 band->checksum = get_bits(&ctx->gb, 16);
305 indx = get_bits(&ctx->gb, 2);
307 av_log(avctx, AV_LOG_ERROR, "Invalid block size!\n");
308 return AVERROR_INVALIDDATA;
310 band->mb_size = 16 >> indx;
311 band->blk_size = 8 >> (indx >> 1);
313 band->inherit_mv = get_bits1(&ctx->gb);
314 band->inherit_qdelta = get_bits1(&ctx->gb);
316 band->glob_quant = get_bits(&ctx->gb, 5);
318 if (!get_bits1(&ctx->gb) || ctx->frame_type == IVI4_FRAMETYPE_INTRA) {
319 transform_id = get_bits(&ctx->gb, 5);
320 if (transform_id >= FF_ARRAY_ELEMS(transforms) ||
321 !transforms[transform_id].inv_trans) {
322 avpriv_request_sample(avctx, "Transform %d", transform_id);
323 return AVERROR_PATCHWELCOME;
325 if ((transform_id >= 7 && transform_id <= 9) ||
326 transform_id == 17) {
327 avpriv_request_sample(avctx, "DCT transform");
328 return AVERROR_PATCHWELCOME;
331 #if IVI4_STREAM_ANALYSER
332 if ((transform_id >= 0 && transform_id <= 2) || transform_id == 10)
336 band->inv_transform = transforms[transform_id].inv_trans;
337 band->dc_transform = transforms[transform_id].dc_trans;
338 band->is_2d_trans = transforms[transform_id].is_2d_trans;
339 if (transform_id < 10)
340 band->transform_size = 8;
342 band->transform_size = 4;
344 if (band->blk_size != band->transform_size)
345 return AVERROR_INVALIDDATA;
347 scan_indx = get_bits(&ctx->gb, 4);
348 if (scan_indx == 15) {
349 av_log(avctx, AV_LOG_ERROR, "Custom scan pattern encountered!\n");
350 return AVERROR_INVALIDDATA;
352 if (scan_indx > 4 && scan_indx < 10) {
353 if (band->blk_size != 4)
354 return AVERROR_INVALIDDATA;
355 } else if (band->blk_size != 8)
356 return AVERROR_INVALIDDATA;
358 band->scan = scan_index_to_tab[scan_indx];
360 band->quant_mat = get_bits(&ctx->gb, 5);
361 if (band->quant_mat >= FF_ARRAY_ELEMS(quant_index_to_tab)) {
363 if (band->quant_mat == 31)
364 av_log(avctx, AV_LOG_ERROR,
365 "Custom quant matrix encountered!\n");
367 avpriv_request_sample(avctx, "Quantization matrix %d",
369 band->quant_mat = -1;
370 return AVERROR_INVALIDDATA;
373 if (old_blk_size != band->blk_size) {
374 av_log(avctx, AV_LOG_ERROR,
375 "The band block size does not match the configuration "
377 return AVERROR_INVALIDDATA;
379 if (band->quant_mat < 0) {
380 av_log(avctx, AV_LOG_ERROR, "Invalid quant_mat inherited\n");
381 return AVERROR_INVALIDDATA;
385 /* decode block huffman codebook */
386 if (!get_bits1(&ctx->gb))
387 band->blk_vlc.tab = ctx->blk_vlc.tab;
389 if (ff_ivi_dec_huff_desc(&ctx->gb, 1, IVI_BLK_HUFF,
390 &band->blk_vlc, avctx))
391 return AVERROR_INVALIDDATA;
393 /* select appropriate rvmap table for this band */
394 band->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
396 /* decode rvmap probability corrections if any */
397 band->num_corr = 0; /* there is no corrections */
398 if (get_bits1(&ctx->gb)) {
399 band->num_corr = get_bits(&ctx->gb, 8); /* get number of correction pairs */
400 if (band->num_corr > 61) {
401 av_log(avctx, AV_LOG_ERROR, "Too many corrections: %d\n",
403 return AVERROR_INVALIDDATA;
406 /* read correction pairs */
407 for (i = 0; i < band->num_corr * 2; i++)
408 band->corr[i] = get_bits(&ctx->gb, 8);
412 if (band->blk_size == 8) {
413 band->intra_base = &ivi4_quant_8x8_intra[quant_index_to_tab[band->quant_mat]][0];
414 band->inter_base = &ivi4_quant_8x8_inter[quant_index_to_tab[band->quant_mat]][0];
416 band->intra_base = &ivi4_quant_4x4_intra[quant_index_to_tab[band->quant_mat]][0];
417 band->inter_base = &ivi4_quant_4x4_inter[quant_index_to_tab[band->quant_mat]][0];
420 /* Indeo 4 doesn't use scale tables */
421 band->intra_scale = NULL;
422 band->inter_scale = NULL;
424 align_get_bits(&ctx->gb);
431 * Decode information (block type, cbp, quant delta, motion vector)
432 * for all macroblocks in the current tile.
434 * @param[in,out] ctx pointer to the decoder context
435 * @param[in,out] band pointer to the band descriptor
436 * @param[in,out] tile pointer to the tile descriptor
437 * @param[in] avctx pointer to the AVCodecContext
438 * @return result code: 0 = OK, negative number = error
440 static int decode_mb_info(IVI45DecContext *ctx, IVIBandDesc *band,
441 IVITile *tile, AVCodecContext *avctx)
443 int x, y, mv_x, mv_y, mv_delta, offs, mb_offset, blks_per_mb,
444 mv_scale, mb_type_bits;
445 IVIMbInfo *mb, *ref_mb;
446 int row_offset = band->mb_size * band->pitch;
449 ref_mb = tile->ref_mbs;
450 offs = tile->ypos * band->pitch + tile->xpos;
452 blks_per_mb = band->mb_size != band->blk_size ? 4 : 1;
453 mb_type_bits = ctx->frame_type == IVI4_FRAMETYPE_BIDIR ? 2 : 1;
455 /* scale factor for motion vectors */
456 mv_scale = (ctx->planes[0].bands[0].mb_size >> 3) - (band->mb_size >> 3);
459 for (y = tile->ypos; y < tile->ypos + tile->height; y += band->mb_size) {
462 for (x = tile->xpos; x < tile->xpos + tile->width; x += band->mb_size) {
465 mb->buf_offs = mb_offset;
469 if (get_bits1(&ctx->gb)) {
470 if (ctx->frame_type == IVI4_FRAMETYPE_INTRA) {
471 av_log(avctx, AV_LOG_ERROR, "Empty macroblock in an INTRA picture!\n");
472 return AVERROR_INVALIDDATA;
474 mb->type = 1; /* empty macroblocks are always INTER */
475 mb->cbp = 0; /* all blocks are empty */
478 if (!band->plane && !band->band_num && ctx->in_q) {
479 mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
481 mb->q_delta = IVI_TOSIGNED(mb->q_delta);
484 mb->mv_x = mb->mv_y = 0; /* no motion vector coded */
485 if (band->inherit_mv && ref_mb) {
486 /* motion vector inheritance */
488 mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
489 mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
491 mb->mv_x = ref_mb->mv_x;
492 mb->mv_y = ref_mb->mv_y;
496 if (band->inherit_mv) {
497 /* copy mb_type from corresponding reference mb */
499 return AVERROR_INVALIDDATA;
500 mb->type = ref_mb->type;
501 } else if (ctx->frame_type == IVI4_FRAMETYPE_INTRA ||
502 ctx->frame_type == IVI4_FRAMETYPE_INTRA1) {
503 mb->type = 0; /* mb_type is always INTRA for intra-frames */
505 mb->type = get_bits(&ctx->gb, mb_type_bits);
508 mb->cbp = get_bits(&ctx->gb, blks_per_mb);
511 if (band->inherit_qdelta) {
512 if (ref_mb) mb->q_delta = ref_mb->q_delta;
513 } else if (mb->cbp || (!band->plane && !band->band_num &&
515 mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
517 mb->q_delta = IVI_TOSIGNED(mb->q_delta);
521 mb->mv_x = mb->mv_y = 0; /* there is no motion vector in intra-macroblocks */
523 if (band->inherit_mv) {
525 /* motion vector inheritance */
527 mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
528 mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
530 mb->mv_x = ref_mb->mv_x;
531 mb->mv_y = ref_mb->mv_y;
534 /* decode motion vector deltas */
535 mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
537 mv_y += IVI_TOSIGNED(mv_delta);
538 mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
540 mv_x += IVI_TOSIGNED(mv_delta);
544 mv_delta = get_vlc2(&ctx->gb,
545 ctx->mb_vlc.tab->table,
547 mv_y += IVI_TOSIGNED(mv_delta);
548 mv_delta = get_vlc2(&ctx->gb,
549 ctx->mb_vlc.tab->table,
551 mv_x += IVI_TOSIGNED(mv_delta);
557 mb->b_mv_x = -mb->mv_x;
558 mb->b_mv_y = -mb->mv_y;
568 mb_offset += band->mb_size;
574 align_get_bits(&ctx->gb);
581 * Rearrange decoding and reference buffers.
583 * @param[in,out] ctx pointer to the decoder context
585 static void switch_buffers(IVI45DecContext *ctx)
587 int is_prev_ref = 0, is_ref = 0;
589 switch (ctx->prev_frame_type) {
590 case IVI4_FRAMETYPE_INTRA:
591 case IVI4_FRAMETYPE_INTRA1:
592 case IVI4_FRAMETYPE_INTER:
597 switch (ctx->frame_type) {
598 case IVI4_FRAMETYPE_INTRA:
599 case IVI4_FRAMETYPE_INTRA1:
600 case IVI4_FRAMETYPE_INTER:
605 if (is_prev_ref && is_ref) {
606 FFSWAP(int, ctx->dst_buf, ctx->ref_buf);
607 } else if (is_prev_ref) {
608 FFSWAP(int, ctx->ref_buf, ctx->b_ref_buf);
609 FFSWAP(int, ctx->dst_buf, ctx->ref_buf);
614 static int is_nonnull_frame(IVI45DecContext *ctx)
616 return ctx->frame_type < IVI4_FRAMETYPE_NULL_FIRST;
620 static av_cold int decode_init(AVCodecContext *avctx)
622 IVI45DecContext *ctx = avctx->priv_data;
624 ff_ivi_init_static_vlc();
626 /* copy rvmap tables in our context so we can apply changes to them */
627 memcpy(ctx->rvmap_tabs, ff_ivi_rvmap_tabs, sizeof(ff_ivi_rvmap_tabs));
629 /* Force allocation of the internal buffers */
630 /* during picture header decoding. */
631 ctx->pic_conf.pic_width = 0;
632 ctx->pic_conf.pic_height = 0;
634 avctx->pix_fmt = AV_PIX_FMT_YUV410P;
636 ctx->decode_pic_hdr = decode_pic_hdr;
637 ctx->decode_band_hdr = decode_band_hdr;
638 ctx->decode_mb_info = decode_mb_info;
639 ctx->switch_buffers = switch_buffers;
640 ctx->is_nonnull_frame = is_nonnull_frame;
646 ctx->b_ref_buf = 3; /* buffer 2 is used for scalability mode */
647 ctx->p_frame = av_frame_alloc();
649 return AVERROR(ENOMEM);
655 AVCodec ff_indeo4_decoder = {
657 .long_name = NULL_IF_CONFIG_SMALL("Intel Indeo Video Interactive 4"),
658 .type = AVMEDIA_TYPE_VIDEO,
659 .id = AV_CODEC_ID_INDEO4,
660 .priv_data_size = sizeof(IVI45DecContext),
662 .close = ff_ivi_decode_close,
663 .decode = ff_ivi_decode_frame,
664 .capabilities = AV_CODEC_CAP_DR1,