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"
37 #define IVI4_PIC_SIZE_ESC 7
41 InvTransformPtr *inv_trans;
42 DCTransformPtr *dc_trans;
45 { ff_ivi_inverse_haar_8x8, ff_ivi_dc_haar_2d, 1 },
46 { ff_ivi_row_haar8, ff_ivi_dc_haar_2d, 0 },
47 { ff_ivi_col_haar8, ff_ivi_dc_haar_2d, 0 },
48 { ff_ivi_put_pixels_8x8, ff_ivi_put_dc_pixel_8x8, 1 },
49 { ff_ivi_inverse_slant_8x8, ff_ivi_dc_slant_2d, 1 },
50 { ff_ivi_row_slant8, ff_ivi_dc_row_slant, 1 },
51 { ff_ivi_col_slant8, ff_ivi_dc_col_slant, 1 },
52 { NULL, NULL, 0 }, /* inverse DCT 8x8 */
53 { NULL, NULL, 0 }, /* inverse DCT 8x1 */
54 { NULL, NULL, 0 }, /* inverse DCT 1x8 */
55 { ff_ivi_inverse_haar_4x4, ff_ivi_dc_haar_2d, 1 },
56 { ff_ivi_inverse_slant_4x4, ff_ivi_dc_slant_2d, 1 },
57 { NULL, NULL, 0 }, /* no transform 4x4 */
58 { ff_ivi_row_haar4, ff_ivi_dc_haar_2d, 0 },
59 { ff_ivi_col_haar4, ff_ivi_dc_haar_2d, 0 },
60 { ff_ivi_row_slant4, ff_ivi_dc_row_slant, 0 },
61 { ff_ivi_col_slant4, ff_ivi_dc_col_slant, 0 },
62 { NULL, NULL, 0 }, /* inverse DCT 4x4 */
66 * Decode subdivision of a plane.
67 * This is a simplified version that checks for two supported subdivisions:
68 * - 1 wavelet band per plane, size factor 1:1, code pattern: 3
69 * - 4 wavelet bands per plane, size factor 1:4, code pattern: 2,3,3,3,3
70 * Anything else is either unsupported or corrupt.
72 * @param[in,out] gb the GetBit context
73 * @return number of wavelet bands or 0 on error
75 static int decode_plane_subdivision(GetBitContext *gb)
79 switch (get_bits(gb, 2)) {
83 for (i = 0; i < 4; i++)
84 if (get_bits(gb, 2) != 3)
92 static inline int scale_tile_size(int def_size, int size_factor)
94 return size_factor == 15 ? def_size : (size_factor + 1) << 5;
98 * Decode Indeo 4 picture header.
100 * @param[in,out] ctx pointer to the decoder context
101 * @param[in] avctx pointer to the AVCodecContext
102 * @return result code: 0 = OK, negative number = error
104 static int decode_pic_hdr(IVI45DecContext *ctx, AVCodecContext *avctx)
106 int pic_size_indx, i, p;
107 IVIPicConfig pic_conf;
109 if (get_bits(&ctx->gb, 18) != 0x3FFF8) {
110 av_log(avctx, AV_LOG_ERROR, "Invalid picture start code!\n");
111 return AVERROR_INVALIDDATA;
114 ctx->prev_frame_type = ctx->frame_type;
115 ctx->frame_type = get_bits(&ctx->gb, 3);
116 if (ctx->frame_type == 7) {
117 av_log(avctx, AV_LOG_ERROR, "Invalid frame type: %d\n", ctx->frame_type);
118 return AVERROR_INVALIDDATA;
121 #if IVI4_STREAM_ANALYSER
122 if (ctx->frame_type == IVI4_FRAMETYPE_BIDIR)
123 ctx->has_b_frames = 1;
126 ctx->transp_status = get_bits1(&ctx->gb);
127 #if IVI4_STREAM_ANALYSER
128 if (ctx->transp_status) {
133 /* unknown bit: Mac decoder ignores this bit, XANIM returns error */
134 if (get_bits1(&ctx->gb)) {
135 av_log(avctx, AV_LOG_ERROR, "Sync bit is set!\n");
136 return AVERROR_INVALIDDATA;
139 ctx->data_size = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 24) : 0;
141 /* null frames don't contain anything else so we just return */
142 if (ctx->frame_type >= IVI4_FRAMETYPE_NULL_FIRST) {
143 av_dlog(avctx, "Null frame encountered!\n");
147 /* Check key lock status. If enabled - ignore lock word. */
148 /* Usually we have to prompt the user for the password, but */
149 /* we don't do that because Indeo 4 videos can be decoded anyway */
150 if (get_bits1(&ctx->gb)) {
151 skip_bits_long(&ctx->gb, 32);
152 av_dlog(avctx, "Password-protected clip!\n");
155 pic_size_indx = get_bits(&ctx->gb, 3);
156 if (pic_size_indx == IVI4_PIC_SIZE_ESC) {
157 pic_conf.pic_height = get_bits(&ctx->gb, 16);
158 pic_conf.pic_width = get_bits(&ctx->gb, 16);
160 pic_conf.pic_height = ivi4_common_pic_sizes[pic_size_indx * 2 + 1];
161 pic_conf.pic_width = ivi4_common_pic_sizes[pic_size_indx * 2 ];
164 /* Decode tile dimensions. */
165 if (get_bits1(&ctx->gb)) {
166 pic_conf.tile_height = scale_tile_size(pic_conf.pic_height, get_bits(&ctx->gb, 4));
167 pic_conf.tile_width = scale_tile_size(pic_conf.pic_width, get_bits(&ctx->gb, 4));
168 #if IVI4_STREAM_ANALYSER
169 ctx->uses_tiling = 1;
172 pic_conf.tile_height = pic_conf.pic_height;
173 pic_conf.tile_width = pic_conf.pic_width;
176 /* Decode chroma subsampling. We support only 4:4 aka YVU9. */
177 if (get_bits(&ctx->gb, 2)) {
178 av_log(avctx, AV_LOG_ERROR, "Only YVU9 picture format is supported!\n");
179 return AVERROR_INVALIDDATA;
181 pic_conf.chroma_height = (pic_conf.pic_height + 3) >> 2;
182 pic_conf.chroma_width = (pic_conf.pic_width + 3) >> 2;
184 /* decode subdivision of the planes */
185 pic_conf.luma_bands = decode_plane_subdivision(&ctx->gb);
186 if (pic_conf.luma_bands)
187 pic_conf.chroma_bands = decode_plane_subdivision(&ctx->gb);
188 ctx->is_scalable = pic_conf.luma_bands != 1 || pic_conf.chroma_bands != 1;
189 if (ctx->is_scalable && (pic_conf.luma_bands != 4 || pic_conf.chroma_bands != 1)) {
190 av_log(avctx, AV_LOG_ERROR, "Scalability: unsupported subdivision! Luma bands: %d, chroma bands: %d\n",
191 pic_conf.luma_bands, pic_conf.chroma_bands);
192 return AVERROR_INVALIDDATA;
195 /* check if picture layout was changed and reallocate buffers */
196 if (ivi_pic_config_cmp(&pic_conf, &ctx->pic_conf)) {
197 if (ff_ivi_init_planes(ctx->planes, &pic_conf, 1)) {
198 av_log(avctx, AV_LOG_ERROR, "Couldn't reallocate color planes!\n");
199 ctx->pic_conf.luma_bands = 0;
200 return AVERROR(ENOMEM);
203 ctx->pic_conf = pic_conf;
205 /* set default macroblock/block dimensions */
206 for (p = 0; p <= 2; p++) {
207 for (i = 0; i < (!p ? pic_conf.luma_bands : pic_conf.chroma_bands); i++) {
208 ctx->planes[p].bands[i].mb_size = !p ? (!ctx->is_scalable ? 16 : 8) : 4;
209 ctx->planes[p].bands[i].blk_size = !p ? 8 : 4;
213 if (ff_ivi_init_tiles(ctx->planes, ctx->pic_conf.tile_width,
214 ctx->pic_conf.tile_height)) {
215 av_log(avctx, AV_LOG_ERROR,
216 "Couldn't reallocate internal structures!\n");
217 return AVERROR(ENOMEM);
221 ctx->frame_num = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 20) : 0;
223 /* skip decTimeEst field if present */
224 if (get_bits1(&ctx->gb))
225 skip_bits(&ctx->gb, 8);
227 /* decode macroblock and block huffman codebooks */
228 if (ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_MB_HUFF, &ctx->mb_vlc, avctx) ||
229 ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_BLK_HUFF, &ctx->blk_vlc, avctx))
230 return AVERROR_INVALIDDATA;
232 ctx->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
234 ctx->in_imf = get_bits1(&ctx->gb);
235 ctx->in_q = get_bits1(&ctx->gb);
237 ctx->pic_glob_quant = get_bits(&ctx->gb, 5);
239 /* TODO: ignore this parameter if unused */
240 ctx->unknown1 = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 0;
242 ctx->checksum = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 16) : 0;
244 /* skip picture header extension if any */
245 while (get_bits1(&ctx->gb)) {
246 av_dlog(avctx, "Pic hdr extension encountered!\n");
247 skip_bits(&ctx->gb, 8);
250 if (get_bits1(&ctx->gb)) {
251 av_log(avctx, AV_LOG_ERROR, "Bad blocks bits encountered!\n");
254 align_get_bits(&ctx->gb);
261 * Decode Indeo 4 band header.
263 * @param[in,out] ctx pointer to the decoder context
264 * @param[in,out] band pointer to the band descriptor
265 * @param[in] avctx pointer to the AVCodecContext
266 * @return result code: 0 = OK, negative number = error
268 static int decode_band_hdr(IVI45DecContext *ctx, IVIBandDesc *band,
269 AVCodecContext *avctx)
271 int plane, band_num, indx, transform_id, scan_indx;
274 plane = get_bits(&ctx->gb, 2);
275 band_num = get_bits(&ctx->gb, 4);
276 if (band->plane != plane || band->band_num != band_num) {
277 av_log(avctx, AV_LOG_ERROR, "Invalid band header sequence!\n");
278 return AVERROR_INVALIDDATA;
281 band->is_empty = get_bits1(&ctx->gb);
282 if (!band->is_empty) {
283 int old_blk_size = band->blk_size;
285 * If header size is not given, header size is 4 bytes. */
286 if (get_bits1(&ctx->gb))
287 skip_bits(&ctx->gb, 16);
289 band->is_halfpel = get_bits(&ctx->gb, 2);
290 if (band->is_halfpel >= 2) {
291 av_log(avctx, AV_LOG_ERROR, "Invalid/unsupported mv resolution: %d!\n",
293 return AVERROR_INVALIDDATA;
295 #if IVI4_STREAM_ANALYSER
296 if (!band->is_halfpel)
297 ctx->uses_fullpel = 1;
300 band->checksum_present = get_bits1(&ctx->gb);
301 if (band->checksum_present)
302 band->checksum = get_bits(&ctx->gb, 16);
304 indx = get_bits(&ctx->gb, 2);
306 av_log(avctx, AV_LOG_ERROR, "Invalid block size!\n");
307 return AVERROR_INVALIDDATA;
309 band->mb_size = 16 >> indx;
310 band->blk_size = 8 >> (indx >> 1);
312 band->inherit_mv = get_bits1(&ctx->gb);
313 band->inherit_qdelta = get_bits1(&ctx->gb);
315 band->glob_quant = get_bits(&ctx->gb, 5);
317 if (!get_bits1(&ctx->gb) || ctx->frame_type == IVI4_FRAMETYPE_INTRA) {
318 transform_id = get_bits(&ctx->gb, 5);
319 if (transform_id >= FF_ARRAY_ELEMS(transforms) ||
320 !transforms[transform_id].inv_trans) {
321 avpriv_request_sample(avctx, "Transform %d", transform_id);
322 return AVERROR_PATCHWELCOME;
324 if ((transform_id >= 7 && transform_id <= 9) ||
325 transform_id == 17) {
326 avpriv_request_sample(avctx, "DCT transform");
327 return AVERROR_PATCHWELCOME;
330 #if IVI4_STREAM_ANALYSER
331 if ((transform_id >= 0 && transform_id <= 2) || transform_id == 10)
335 band->inv_transform = transforms[transform_id].inv_trans;
336 band->dc_transform = transforms[transform_id].dc_trans;
337 band->is_2d_trans = transforms[transform_id].is_2d_trans;
338 if (transform_id < 10)
339 band->transform_size = 8;
341 band->transform_size = 4;
343 if (band->blk_size != band->transform_size)
344 return AVERROR_INVALIDDATA;
346 scan_indx = get_bits(&ctx->gb, 4);
347 if (scan_indx == 15) {
348 av_log(avctx, AV_LOG_ERROR, "Custom scan pattern encountered!\n");
349 return AVERROR_INVALIDDATA;
351 if (scan_indx > 4 && scan_indx < 10) {
352 if (band->blk_size != 4)
353 return AVERROR_INVALIDDATA;
354 } else if (band->blk_size != 8)
355 return AVERROR_INVALIDDATA;
357 band->scan = scan_index_to_tab[scan_indx];
359 band->quant_mat = get_bits(&ctx->gb, 5);
360 if (band->quant_mat >= FF_ARRAY_ELEMS(quant_index_to_tab)) {
362 if (band->quant_mat == 31)
363 av_log(avctx, AV_LOG_ERROR,
364 "Custom quant matrix encountered!\n");
366 avpriv_request_sample(avctx, "Quantization matrix %d",
368 band->quant_mat = -1;
369 return AVERROR_INVALIDDATA;
372 if (old_blk_size != band->blk_size) {
373 av_log(avctx, AV_LOG_ERROR,
374 "The band block size does not match the configuration "
376 return AVERROR_INVALIDDATA;
378 if (band->quant_mat < 0) {
379 av_log(avctx, AV_LOG_ERROR, "Invalid quant_mat inherited\n");
380 return AVERROR_INVALIDDATA;
384 /* decode block huffman codebook */
385 if (!get_bits1(&ctx->gb))
386 band->blk_vlc.tab = ctx->blk_vlc.tab;
388 if (ff_ivi_dec_huff_desc(&ctx->gb, 1, IVI_BLK_HUFF,
389 &band->blk_vlc, avctx))
390 return AVERROR_INVALIDDATA;
392 /* select appropriate rvmap table for this band */
393 band->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
395 /* decode rvmap probability corrections if any */
396 band->num_corr = 0; /* there is no corrections */
397 if (get_bits1(&ctx->gb)) {
398 band->num_corr = get_bits(&ctx->gb, 8); /* get number of correction pairs */
399 if (band->num_corr > 61) {
400 av_log(avctx, AV_LOG_ERROR, "Too many corrections: %d\n",
402 return AVERROR_INVALIDDATA;
405 /* read correction pairs */
406 for (i = 0; i < band->num_corr * 2; i++)
407 band->corr[i] = get_bits(&ctx->gb, 8);
411 if (band->blk_size == 8) {
412 band->intra_base = &ivi4_quant_8x8_intra[quant_index_to_tab[band->quant_mat]][0];
413 band->inter_base = &ivi4_quant_8x8_inter[quant_index_to_tab[band->quant_mat]][0];
415 band->intra_base = &ivi4_quant_4x4_intra[quant_index_to_tab[band->quant_mat]][0];
416 band->inter_base = &ivi4_quant_4x4_inter[quant_index_to_tab[band->quant_mat]][0];
419 /* Indeo 4 doesn't use scale tables */
420 band->intra_scale = NULL;
421 band->inter_scale = NULL;
423 align_get_bits(&ctx->gb);
430 * Decode information (block type, cbp, quant delta, motion vector)
431 * for all macroblocks in the current tile.
433 * @param[in,out] ctx pointer to the decoder context
434 * @param[in,out] band pointer to the band descriptor
435 * @param[in,out] tile pointer to the tile descriptor
436 * @param[in] avctx pointer to the AVCodecContext
437 * @return result code: 0 = OK, negative number = error
439 static int decode_mb_info(IVI45DecContext *ctx, IVIBandDesc *band,
440 IVITile *tile, AVCodecContext *avctx)
442 int x, y, mv_x, mv_y, mv_delta, offs, mb_offset, blks_per_mb,
443 mv_scale, mb_type_bits;
444 IVIMbInfo *mb, *ref_mb;
445 int row_offset = band->mb_size * band->pitch;
448 ref_mb = tile->ref_mbs;
449 offs = tile->ypos * band->pitch + tile->xpos;
451 blks_per_mb = band->mb_size != band->blk_size ? 4 : 1;
452 mb_type_bits = ctx->frame_type == IVI4_FRAMETYPE_BIDIR ? 2 : 1;
454 /* scale factor for motion vectors */
455 mv_scale = (ctx->planes[0].bands[0].mb_size >> 3) - (band->mb_size >> 3);
458 for (y = tile->ypos; y < tile->ypos + tile->height; y += band->mb_size) {
461 for (x = tile->xpos; x < tile->xpos + tile->width; x += band->mb_size) {
464 mb->buf_offs = mb_offset;
468 if (get_bits1(&ctx->gb)) {
469 if (ctx->frame_type == IVI4_FRAMETYPE_INTRA) {
470 av_log(avctx, AV_LOG_ERROR, "Empty macroblock in an INTRA picture!\n");
471 return AVERROR_INVALIDDATA;
473 mb->type = 1; /* empty macroblocks are always INTER */
474 mb->cbp = 0; /* all blocks are empty */
477 if (!band->plane && !band->band_num && ctx->in_q) {
478 mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
480 mb->q_delta = IVI_TOSIGNED(mb->q_delta);
483 mb->mv_x = mb->mv_y = 0; /* no motion vector coded */
484 if (band->inherit_mv && ref_mb) {
485 /* motion vector inheritance */
487 mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
488 mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
490 mb->mv_x = ref_mb->mv_x;
491 mb->mv_y = ref_mb->mv_y;
495 if (band->inherit_mv) {
496 /* copy mb_type from corresponding reference mb */
498 return AVERROR_INVALIDDATA;
499 mb->type = ref_mb->type;
500 } else if (ctx->frame_type == IVI4_FRAMETYPE_INTRA ||
501 ctx->frame_type == IVI4_FRAMETYPE_INTRA1) {
502 mb->type = 0; /* mb_type is always INTRA for intra-frames */
504 mb->type = get_bits(&ctx->gb, mb_type_bits);
507 mb->cbp = get_bits(&ctx->gb, blks_per_mb);
510 if (band->inherit_qdelta) {
511 if (ref_mb) mb->q_delta = ref_mb->q_delta;
512 } else if (mb->cbp || (!band->plane && !band->band_num &&
514 mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
516 mb->q_delta = IVI_TOSIGNED(mb->q_delta);
520 mb->mv_x = mb->mv_y = 0; /* there is no motion vector in intra-macroblocks */
522 if (band->inherit_mv) {
524 /* motion vector inheritance */
526 mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
527 mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
529 mb->mv_x = ref_mb->mv_x;
530 mb->mv_y = ref_mb->mv_y;
533 /* decode motion vector deltas */
534 mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
536 mv_y += IVI_TOSIGNED(mv_delta);
537 mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
539 mv_x += IVI_TOSIGNED(mv_delta);
543 mv_delta = get_vlc2(&ctx->gb,
544 ctx->mb_vlc.tab->table,
546 mv_y += IVI_TOSIGNED(mv_delta);
547 mv_delta = get_vlc2(&ctx->gb,
548 ctx->mb_vlc.tab->table,
550 mv_x += IVI_TOSIGNED(mv_delta);
556 mb->b_mv_x = -mb->mv_x;
557 mb->b_mv_y = -mb->mv_y;
567 mb_offset += band->mb_size;
573 align_get_bits(&ctx->gb);
580 * Rearrange decoding and reference buffers.
582 * @param[in,out] ctx pointer to the decoder context
584 static void switch_buffers(IVI45DecContext *ctx)
586 int is_prev_ref = 0, is_ref = 0;
588 switch (ctx->prev_frame_type) {
589 case IVI4_FRAMETYPE_INTRA:
590 case IVI4_FRAMETYPE_INTRA1:
591 case IVI4_FRAMETYPE_INTER:
596 switch (ctx->frame_type) {
597 case IVI4_FRAMETYPE_INTRA:
598 case IVI4_FRAMETYPE_INTRA1:
599 case IVI4_FRAMETYPE_INTER:
604 if (is_prev_ref && is_ref) {
605 FFSWAP(int, ctx->dst_buf, ctx->ref_buf);
606 } else if (is_prev_ref) {
607 FFSWAP(int, ctx->ref_buf, ctx->b_ref_buf);
608 FFSWAP(int, ctx->dst_buf, ctx->ref_buf);
613 static int is_nonnull_frame(IVI45DecContext *ctx)
615 return ctx->frame_type < IVI4_FRAMETYPE_NULL_FIRST;
619 static av_cold int decode_init(AVCodecContext *avctx)
621 IVI45DecContext *ctx = avctx->priv_data;
623 ff_ivi_init_static_vlc();
625 /* copy rvmap tables in our context so we can apply changes to them */
626 memcpy(ctx->rvmap_tabs, ff_ivi_rvmap_tabs, sizeof(ff_ivi_rvmap_tabs));
628 /* Force allocation of the internal buffers */
629 /* during picture header decoding. */
630 ctx->pic_conf.pic_width = 0;
631 ctx->pic_conf.pic_height = 0;
633 avctx->pix_fmt = AV_PIX_FMT_YUV410P;
635 ctx->decode_pic_hdr = decode_pic_hdr;
636 ctx->decode_band_hdr = decode_band_hdr;
637 ctx->decode_mb_info = decode_mb_info;
638 ctx->switch_buffers = switch_buffers;
639 ctx->is_nonnull_frame = is_nonnull_frame;
645 ctx->b_ref_buf = 3; /* buffer 2 is used for scalability mode */
646 ctx->p_frame = av_frame_alloc();
648 return AVERROR(ENOMEM);
654 AVCodec ff_indeo4_decoder = {
656 .long_name = NULL_IF_CONFIG_SMALL("Intel Indeo Video Interactive 4"),
657 .type = AVMEDIA_TYPE_VIDEO,
658 .id = AV_CODEC_ID_INDEO4,
659 .priv_data_size = sizeof(IVI45DecContext),
661 .close = ff_ivi_decode_close,
662 .decode = ff_ivi_decode_frame,
663 .capabilities = CODEC_CAP_DR1,