You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

874 lines
31KB

  1. /*
  2. * Indeo Video Interactive v4 compatible decoder
  3. * Copyright (c) 2009-2011 Maxim Poliakovski
  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. * @file
  23. * Indeo Video Interactive version 4 decoder
  24. *
  25. * Indeo 4 data is usually transported within .avi or .mov files.
  26. * Known FOURCCs: 'IV41'
  27. */
  28. #define BITSTREAM_READER_LE
  29. #include "avcodec.h"
  30. #include "get_bits.h"
  31. #include "dsputil.h"
  32. #include "ivi_dsp.h"
  33. #include "ivi_common.h"
  34. #include "indeo4data.h"
  35. #define IVI4_STREAM_ANALYSER 0
  36. #define IVI4_DEBUG_CHECKSUM 0
  37. /**
  38. * Indeo 4 frame types.
  39. */
  40. enum {
  41. FRAMETYPE_INTRA = 0,
  42. FRAMETYPE_BIDIR1 = 1, ///< bidirectional frame
  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
  48. };
  49. #define IVI4_PIC_SIZE_ESC 7
  50. typedef struct {
  51. GetBitContext gb;
  52. AVFrame frame;
  53. RVMapDesc rvmap_tabs[9]; ///< local corrected copy of the static rvmap tables
  54. uint32_t frame_num;
  55. int frame_type;
  56. int prev_frame_type; ///< frame type of the previous frame
  57. uint32_t data_size; ///< size of the frame data in bytes from picture header
  58. int is_scalable;
  59. int transp_status; ///< transparency mode status: 1 - enabled
  60. IVIPicConfig pic_conf;
  61. IVIPlaneDesc planes[3]; ///< color planes
  62. int buf_switch; ///< used to switch between three buffers
  63. int dst_buf; ///< buffer index for the currently decoded frame
  64. int ref_buf; ///< inter frame reference buffer index
  65. IVIHuffTab mb_vlc; ///< current macroblock table descriptor
  66. IVIHuffTab blk_vlc; ///< current block table descriptor
  67. uint16_t checksum; ///< frame checksum
  68. uint8_t rvmap_sel;
  69. uint8_t in_imf;
  70. uint8_t in_q; ///< flag for explicitly stored quantiser delta
  71. uint8_t pic_glob_quant;
  72. uint8_t unknown1;
  73. #if IVI4_STREAM_ANALYSER
  74. uint8_t has_b_frames;
  75. uint8_t has_transp;
  76. uint8_t uses_tiling;
  77. uint8_t uses_haar;
  78. uint8_t uses_fullpel;
  79. #endif
  80. } IVI4DecContext;
  81. static const struct {
  82. InvTransformPtr *inv_trans;
  83. DCTransformPtr *dc_trans;
  84. int is_2d_trans;
  85. } transforms[18] = {
  86. { ff_ivi_inverse_haar_8x8, ff_ivi_dc_haar_2d, 1 },
  87. { NULL, NULL, 0 }, /* inverse Haar 8x1 */
  88. { NULL, NULL, 0 }, /* inverse Haar 1x8 */
  89. { ff_ivi_put_pixels_8x8, ff_ivi_put_dc_pixel_8x8, 1 },
  90. { ff_ivi_inverse_slant_8x8, ff_ivi_dc_slant_2d, 1 },
  91. { ff_ivi_row_slant8, ff_ivi_dc_row_slant, 1 },
  92. { ff_ivi_col_slant8, ff_ivi_dc_col_slant, 1 },
  93. { NULL, NULL, 0 }, /* inverse DCT 8x8 */
  94. { NULL, NULL, 0 }, /* inverse DCT 8x1 */
  95. { NULL, NULL, 0 }, /* inverse DCT 1x8 */
  96. { NULL, NULL, 0 }, /* inverse Haar 4x4 */
  97. { ff_ivi_inverse_slant_4x4, ff_ivi_dc_slant_2d, 1 },
  98. { NULL, NULL, 0 }, /* no transform 4x4 */
  99. { NULL, NULL, 0 }, /* inverse Haar 1x4 */
  100. { NULL, NULL, 0 }, /* inverse Haar 4x1 */
  101. { NULL, NULL, 0 }, /* inverse slant 1x4 */
  102. { NULL, NULL, 0 }, /* inverse slant 4x1 */
  103. { NULL, NULL, 0 }, /* inverse DCT 4x4 */
  104. };
  105. /**
  106. * Decode subdivision of a plane.
  107. * This is a simplified version that checks for two supported subdivisions:
  108. * - 1 wavelet band per plane, size factor 1:1, code pattern: 3
  109. * - 4 wavelet bands per plane, size factor 1:4, code pattern: 2,3,3,3,3
  110. * Anything else is either unsupported or corrupt.
  111. *
  112. * @param[in,out] gb the GetBit context
  113. * @return number of wavelet bands or 0 on error
  114. */
  115. static int decode_plane_subdivision(GetBitContext *gb)
  116. {
  117. int i;
  118. switch (get_bits(gb, 2)) {
  119. case 3:
  120. return 1;
  121. case 2:
  122. for (i = 0; i < 4; i++)
  123. if (get_bits(gb, 2) != 3)
  124. return 0;
  125. return 4;
  126. default:
  127. return 0;
  128. }
  129. }
  130. static inline int scale_tile_size(int def_size, int size_factor)
  131. {
  132. return size_factor == 15 ? def_size : (size_factor + 1) << 5;
  133. }
  134. /**
  135. * Decode Indeo 4 picture header.
  136. *
  137. * @param[in,out] ctx pointer to the decoder context
  138. * @param[in] avctx pointer to the AVCodecContext
  139. * @return result code: 0 = OK, negative number = error
  140. */
  141. static int decode_pic_hdr(IVI4DecContext *ctx, AVCodecContext *avctx)
  142. {
  143. int pic_size_indx, i, p;
  144. IVIPicConfig pic_conf;
  145. if (get_bits(&ctx->gb, 18) != 0x3FFF8) {
  146. av_log(avctx, AV_LOG_ERROR, "Invalid picture start code!\n");
  147. return AVERROR_INVALIDDATA;
  148. }
  149. ctx->prev_frame_type = ctx->frame_type;
  150. ctx->frame_type = get_bits(&ctx->gb, 3);
  151. if (ctx->frame_type == 7) {
  152. av_log(avctx, AV_LOG_ERROR, "Invalid frame type: %d\n", ctx->frame_type);
  153. return AVERROR_INVALIDDATA;
  154. }
  155. #if IVI4_STREAM_ANALYSER
  156. if ( ctx->frame_type == FRAMETYPE_BIDIR1
  157. || ctx->frame_type == FRAMETYPE_BIDIR)
  158. ctx->has_b_frames = 1;
  159. #endif
  160. ctx->transp_status = get_bits1(&ctx->gb);
  161. #if IVI4_STREAM_ANALYSER
  162. if (ctx->transp_status) {
  163. ctx->has_transp = 1;
  164. }
  165. #endif
  166. /* unknown bit: Mac decoder ignores this bit, XANIM returns error */
  167. if (get_bits1(&ctx->gb)) {
  168. av_log(avctx, AV_LOG_ERROR, "Sync bit is set!\n");
  169. return AVERROR_INVALIDDATA;
  170. }
  171. ctx->data_size = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 24) : 0;
  172. /* null frames don't contain anything else so we just return */
  173. if (ctx->frame_type >= FRAMETYPE_NULL_FIRST) {
  174. av_dlog(avctx, "Null frame encountered!\n");
  175. return 0;
  176. }
  177. /* Check key lock status. If enabled - ignore lock word. */
  178. /* Usually we have to prompt the user for the password, but */
  179. /* we don't do that because Indeo 4 videos can be decoded anyway */
  180. if (get_bits1(&ctx->gb)) {
  181. skip_bits_long(&ctx->gb, 32);
  182. av_dlog(avctx, "Password-protected clip!\n");
  183. }
  184. pic_size_indx = get_bits(&ctx->gb, 3);
  185. if (pic_size_indx == IVI4_PIC_SIZE_ESC) {
  186. pic_conf.pic_height = get_bits(&ctx->gb, 16);
  187. pic_conf.pic_width = get_bits(&ctx->gb, 16);
  188. } else {
  189. pic_conf.pic_height = ivi4_common_pic_sizes[pic_size_indx * 2 + 1];
  190. pic_conf.pic_width = ivi4_common_pic_sizes[pic_size_indx * 2 ];
  191. }
  192. /* Decode tile dimensions. */
  193. if (get_bits1(&ctx->gb)) {
  194. pic_conf.tile_height = scale_tile_size(pic_conf.pic_height, get_bits(&ctx->gb, 4));
  195. pic_conf.tile_width = scale_tile_size(pic_conf.pic_width, get_bits(&ctx->gb, 4));
  196. #if IVI4_STREAM_ANALYSER
  197. ctx->uses_tiling = 1;
  198. #endif
  199. } else {
  200. pic_conf.tile_height = pic_conf.pic_height;
  201. pic_conf.tile_width = pic_conf.pic_width;
  202. }
  203. /* Decode chroma subsampling. We support only 4:4 aka YVU9. */
  204. if (get_bits(&ctx->gb, 2)) {
  205. av_log(avctx, AV_LOG_ERROR, "Only YVU9 picture format is supported!\n");
  206. return AVERROR_INVALIDDATA;
  207. }
  208. pic_conf.chroma_height = (pic_conf.pic_height + 3) >> 2;
  209. pic_conf.chroma_width = (pic_conf.pic_width + 3) >> 2;
  210. /* decode subdivision of the planes */
  211. pic_conf.luma_bands = decode_plane_subdivision(&ctx->gb);
  212. if (pic_conf.luma_bands)
  213. pic_conf.chroma_bands = decode_plane_subdivision(&ctx->gb);
  214. ctx->is_scalable = pic_conf.luma_bands != 1 || pic_conf.chroma_bands != 1;
  215. if (ctx->is_scalable && (pic_conf.luma_bands != 4 || pic_conf.chroma_bands != 1)) {
  216. av_log(avctx, AV_LOG_ERROR, "Scalability: unsupported subdivision! Luma bands: %d, chroma bands: %d\n",
  217. pic_conf.luma_bands, pic_conf.chroma_bands);
  218. return AVERROR_INVALIDDATA;
  219. }
  220. /* check if picture layout was changed and reallocate buffers */
  221. if (ivi_pic_config_cmp(&pic_conf, &ctx->pic_conf)) {
  222. if (ff_ivi_init_planes(ctx->planes, &pic_conf)) {
  223. av_log(avctx, AV_LOG_ERROR, "Couldn't reallocate color planes!\n");
  224. return AVERROR(ENOMEM);
  225. }
  226. ctx->pic_conf = pic_conf;
  227. /* set default macroblock/block dimensions */
  228. for (p = 0; p <= 2; p++) {
  229. for (i = 0; i < (!p ? pic_conf.luma_bands : pic_conf.chroma_bands); i++) {
  230. ctx->planes[p].bands[i].mb_size = !p ? (!ctx->is_scalable ? 16 : 8) : 4;
  231. ctx->planes[p].bands[i].blk_size = !p ? 8 : 4;
  232. }
  233. }
  234. if (ff_ivi_init_tiles(ctx->planes, ctx->pic_conf.tile_width,
  235. ctx->pic_conf.tile_height)) {
  236. av_log(avctx, AV_LOG_ERROR,
  237. "Couldn't reallocate internal structures!\n");
  238. return AVERROR(ENOMEM);
  239. }
  240. }
  241. ctx->frame_num = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 20) : 0;
  242. /* skip decTimeEst field if present */
  243. if (get_bits1(&ctx->gb))
  244. skip_bits(&ctx->gb, 8);
  245. /* decode macroblock and block huffman codebooks */
  246. if (ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_MB_HUFF, &ctx->mb_vlc, avctx) ||
  247. ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_BLK_HUFF, &ctx->blk_vlc, avctx))
  248. return AVERROR_INVALIDDATA;
  249. ctx->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
  250. ctx->in_imf = get_bits1(&ctx->gb);
  251. ctx->in_q = get_bits1(&ctx->gb);
  252. ctx->pic_glob_quant = get_bits(&ctx->gb, 5);
  253. /* TODO: ignore this parameter if unused */
  254. 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;
  256. /* skip picture header extension if any */
  257. while (get_bits1(&ctx->gb)) {
  258. av_dlog(avctx, "Pic hdr extension encountered!\n");
  259. skip_bits(&ctx->gb, 8);
  260. }
  261. if (get_bits1(&ctx->gb)) {
  262. av_log(avctx, AV_LOG_ERROR, "Bad blocks bits encountered!\n");
  263. }
  264. align_get_bits(&ctx->gb);
  265. return 0;
  266. }
  267. /**
  268. * Decode Indeo 4 band header.
  269. *
  270. * @param[in,out] ctx pointer to the decoder context
  271. * @param[in,out] band pointer to the band descriptor
  272. * @param[in] avctx pointer to the AVCodecContext
  273. * @return result code: 0 = OK, negative number = error
  274. */
  275. static int decode_band_hdr(IVI4DecContext *ctx, IVIBandDesc *band,
  276. AVCodecContext *avctx)
  277. {
  278. int plane, band_num, indx, transform_id, scan_indx;
  279. int i;
  280. plane = get_bits(&ctx->gb, 2);
  281. band_num = get_bits(&ctx->gb, 4);
  282. if (band->plane != plane || band->band_num != band_num) {
  283. av_log(avctx, AV_LOG_ERROR, "Invalid band header sequence!\n");
  284. return AVERROR_INVALIDDATA;
  285. }
  286. band->is_empty = get_bits1(&ctx->gb);
  287. if (!band->is_empty) {
  288. /* skip header size
  289. * If header size is not given, header size is 4 bytes. */
  290. if (get_bits1(&ctx->gb))
  291. skip_bits(&ctx->gb, 16);
  292. band->is_halfpel = get_bits(&ctx->gb, 2);
  293. if (band->is_halfpel >= 2) {
  294. av_log(avctx, AV_LOG_ERROR, "Invalid/unsupported mv resolution: %d!\n",
  295. band->is_halfpel);
  296. return AVERROR_INVALIDDATA;
  297. }
  298. #if IVI4_STREAM_ANALYSER
  299. if (!band->is_halfpel)
  300. ctx->uses_fullpel = 1;
  301. #endif
  302. band->checksum_present = get_bits1(&ctx->gb);
  303. if (band->checksum_present)
  304. band->checksum = get_bits(&ctx->gb, 16);
  305. indx = get_bits(&ctx->gb, 2);
  306. if (indx == 3) {
  307. av_log(avctx, AV_LOG_ERROR, "Invalid block size!\n");
  308. return AVERROR_INVALIDDATA;
  309. }
  310. band->mb_size = 16 >> indx;
  311. band->blk_size = 8 >> (indx >> 1);
  312. band->inherit_mv = get_bits1(&ctx->gb);
  313. band->inherit_qdelta = get_bits1(&ctx->gb);
  314. band->glob_quant = get_bits(&ctx->gb, 5);
  315. if (!get_bits1(&ctx->gb) || ctx->frame_type == FRAMETYPE_INTRA) {
  316. transform_id = get_bits(&ctx->gb, 5);
  317. if (transform_id >= FF_ARRAY_ELEMS(transforms) ||
  318. !transforms[transform_id].inv_trans) {
  319. av_log_ask_for_sample(avctx, "Unimplemented transform: %d!\n", transform_id);
  320. return AVERROR_PATCHWELCOME;
  321. }
  322. if ((transform_id >= 7 && transform_id <= 9) ||
  323. transform_id == 17) {
  324. av_log_ask_for_sample(avctx, "DCT transform not supported yet!\n");
  325. return AVERROR_PATCHWELCOME;
  326. }
  327. if (transform_id < 10 && band->blk_size < 8) {
  328. av_log(avctx, AV_LOG_ERROR, "wrong transform size!\n");
  329. return AVERROR_INVALIDDATA;
  330. }
  331. #if IVI4_STREAM_ANALYSER
  332. if ((transform_id >= 0 && transform_id <= 2) || transform_id == 10)
  333. ctx->uses_haar = 1;
  334. #endif
  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. band->transform_size= (transform_id < 10) ? 8 : 4;
  339. scan_indx = get_bits(&ctx->gb, 4);
  340. if ((scan_indx>4 && scan_indx<10) != (band->blk_size==4)) {
  341. av_log(avctx, AV_LOG_ERROR, "mismatching scan table!\n");
  342. return AVERROR_INVALIDDATA;
  343. }
  344. if (scan_indx == 15) {
  345. av_log(avctx, AV_LOG_ERROR, "Custom scan pattern encountered!\n");
  346. return AVERROR_INVALIDDATA;
  347. }
  348. band->scan = scan_index_to_tab[scan_indx];
  349. band->quant_mat = get_bits(&ctx->gb, 5);
  350. if (band->quant_mat == 31) {
  351. av_log(avctx, AV_LOG_ERROR, "Custom quant matrix encountered!\n");
  352. return AVERROR_INVALIDDATA;
  353. }
  354. if (band->quant_mat > 21) {
  355. av_log(avctx, AV_LOG_ERROR, "Invalid quant matrix encountered!\n");
  356. return AVERROR_INVALIDDATA;
  357. }
  358. }
  359. /* decode block huffman codebook */
  360. if (ff_ivi_dec_huff_desc(&ctx->gb, get_bits1(&ctx->gb), IVI_BLK_HUFF,
  361. &band->blk_vlc, avctx))
  362. return AVERROR_INVALIDDATA;
  363. /* select appropriate rvmap table for this band */
  364. band->rvmap_sel = get_bits1(&ctx->gb) ? get_bits(&ctx->gb, 3) : 8;
  365. /* decode rvmap probability corrections if any */
  366. band->num_corr = 0; /* there is no corrections */
  367. if (get_bits1(&ctx->gb)) {
  368. band->num_corr = get_bits(&ctx->gb, 8); /* get number of correction pairs */
  369. if (band->num_corr > 61) {
  370. av_log(avctx, AV_LOG_ERROR, "Too many corrections: %d\n",
  371. band->num_corr);
  372. return AVERROR_INVALIDDATA;
  373. }
  374. /* read correction pairs */
  375. for (i = 0; i < band->num_corr * 2; i++)
  376. band->corr[i] = get_bits(&ctx->gb, 8);
  377. }
  378. }
  379. if (band->blk_size == 8) {
  380. band->intra_base = &ivi4_quant_8x8_intra[quant_index_to_tab[band->quant_mat]][0];
  381. band->inter_base = &ivi4_quant_8x8_inter[quant_index_to_tab[band->quant_mat]][0];
  382. } else {
  383. band->intra_base = &ivi4_quant_4x4_intra[quant_index_to_tab[band->quant_mat]][0];
  384. band->inter_base = &ivi4_quant_4x4_inter[quant_index_to_tab[band->quant_mat]][0];
  385. }
  386. /* Indeo 4 doesn't use scale tables */
  387. band->intra_scale = NULL;
  388. band->inter_scale = NULL;
  389. align_get_bits(&ctx->gb);
  390. if (!band->scan) {
  391. av_log(avctx, AV_LOG_ERROR, "band->scan not set\n");
  392. return AVERROR_INVALIDDATA;
  393. }
  394. return 0;
  395. }
  396. /**
  397. * Decode information (block type, cbp, quant delta, motion vector)
  398. * for all macroblocks in the current tile.
  399. *
  400. * @param[in,out] ctx pointer to the decoder context
  401. * @param[in,out] band pointer to the band descriptor
  402. * @param[in,out] tile pointer to the tile descriptor
  403. * @param[in] avctx pointer to the AVCodecContext
  404. * @return result code: 0 = OK, negative number = error
  405. */
  406. static int decode_mb_info(IVI4DecContext *ctx, IVIBandDesc *band,
  407. IVITile *tile, AVCodecContext *avctx)
  408. {
  409. int x, y, mv_x, mv_y, mv_delta, offs, mb_offset, blks_per_mb,
  410. mv_scale, mb_type_bits, s;
  411. IVIMbInfo *mb, *ref_mb;
  412. int row_offset = band->mb_size * band->pitch;
  413. mb = tile->mbs;
  414. ref_mb = tile->ref_mbs;
  415. offs = tile->ypos * band->pitch + tile->xpos;
  416. blks_per_mb = band->mb_size != band->blk_size ? 4 : 1;
  417. mb_type_bits = ctx->frame_type == FRAMETYPE_BIDIR ? 2 : 1;
  418. /* scale factor for motion vectors */
  419. mv_scale = (ctx->planes[0].bands[0].mb_size >> 3) - (band->mb_size >> 3);
  420. mv_x = mv_y = 0;
  421. if (((tile->width + band->mb_size-1)/band->mb_size) * ((tile->height + band->mb_size-1)/band->mb_size) != tile->num_MBs) {
  422. av_log(avctx, AV_LOG_ERROR, "num_MBs mismatch %d %d %d %d\n", tile->width, tile->height, band->mb_size, tile->num_MBs);
  423. return -1;
  424. }
  425. for (y = tile->ypos; y < tile->ypos + tile->height; y += band->mb_size) {
  426. mb_offset = offs;
  427. for (x = tile->xpos; x < tile->xpos + tile->width; x += band->mb_size) {
  428. mb->xpos = x;
  429. mb->ypos = y;
  430. mb->buf_offs = mb_offset;
  431. if (get_bits1(&ctx->gb)) {
  432. if (ctx->frame_type == FRAMETYPE_INTRA) {
  433. av_log(avctx, AV_LOG_ERROR, "Empty macroblock in an INTRA picture!\n");
  434. return AVERROR_INVALIDDATA;
  435. }
  436. mb->type = 1; /* empty macroblocks are always INTER */
  437. mb->cbp = 0; /* all blocks are empty */
  438. mb->q_delta = 0;
  439. if (!band->plane && !band->band_num && ctx->in_q) {
  440. mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
  441. IVI_VLC_BITS, 1);
  442. mb->q_delta = IVI_TOSIGNED(mb->q_delta);
  443. }
  444. mb->mv_x = mb->mv_y = 0; /* no motion vector coded */
  445. if (band->inherit_mv && ref_mb) {
  446. /* motion vector inheritance */
  447. if (mv_scale) {
  448. mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
  449. mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
  450. } else {
  451. mb->mv_x = ref_mb->mv_x;
  452. mb->mv_y = ref_mb->mv_y;
  453. }
  454. }
  455. } else {
  456. if (band->inherit_mv && ref_mb) {
  457. mb->type = ref_mb->type; /* copy mb_type from corresponding reference mb */
  458. } else if (ctx->frame_type == FRAMETYPE_INTRA) {
  459. mb->type = 0; /* mb_type is always INTRA for intra-frames */
  460. } else {
  461. mb->type = get_bits(&ctx->gb, mb_type_bits);
  462. }
  463. mb->cbp = get_bits(&ctx->gb, blks_per_mb);
  464. mb->q_delta = 0;
  465. if (band->inherit_qdelta) {
  466. if (ref_mb) mb->q_delta = ref_mb->q_delta;
  467. } else if (mb->cbp || (!band->plane && !band->band_num &&
  468. ctx->in_q)) {
  469. mb->q_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
  470. IVI_VLC_BITS, 1);
  471. mb->q_delta = IVI_TOSIGNED(mb->q_delta);
  472. }
  473. if (!mb->type) {
  474. mb->mv_x = mb->mv_y = 0; /* there is no motion vector in intra-macroblocks */
  475. } else {
  476. if (band->inherit_mv && ref_mb) {
  477. /* motion vector inheritance */
  478. if (mv_scale) {
  479. mb->mv_x = ivi_scale_mv(ref_mb->mv_x, mv_scale);
  480. mb->mv_y = ivi_scale_mv(ref_mb->mv_y, mv_scale);
  481. } else {
  482. mb->mv_x = ref_mb->mv_x;
  483. mb->mv_y = ref_mb->mv_y;
  484. }
  485. } else {
  486. /* decode motion vector deltas */
  487. mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
  488. IVI_VLC_BITS, 1);
  489. mv_y += IVI_TOSIGNED(mv_delta);
  490. mv_delta = get_vlc2(&ctx->gb, ctx->mb_vlc.tab->table,
  491. IVI_VLC_BITS, 1);
  492. mv_x += IVI_TOSIGNED(mv_delta);
  493. mb->mv_x = mv_x;
  494. mb->mv_y = mv_y;
  495. }
  496. }
  497. }
  498. s= band->is_halfpel;
  499. if (mb->type)
  500. if ( x + (mb->mv_x >>s) + (y+ (mb->mv_y >>s))*band->pitch < 0 ||
  501. x + ((mb->mv_x+s)>>s) + band->mb_size - 1
  502. + (y+band->mb_size - 1 +((mb->mv_y+s)>>s))*band->pitch > band->bufsize -1) {
  503. av_log(avctx, AV_LOG_ERROR, "motion vector %d %d outside reference\n", x*s + mb->mv_x, y*s + mb->mv_y);
  504. return AVERROR_INVALIDDATA;
  505. }
  506. mb++;
  507. if (ref_mb)
  508. ref_mb++;
  509. mb_offset += band->mb_size;
  510. }
  511. offs += row_offset;
  512. }
  513. align_get_bits(&ctx->gb);
  514. return 0;
  515. }
  516. /**
  517. * Decode an Indeo 4 band.
  518. *
  519. * @param[in,out] ctx pointer to the decoder context
  520. * @param[in,out] band pointer to the band descriptor
  521. * @param[in] avctx pointer to the AVCodecContext
  522. * @return result code: 0 = OK, negative number = error
  523. */
  524. static int decode_band(IVI4DecContext *ctx, int plane_num,
  525. IVIBandDesc *band, AVCodecContext *avctx)
  526. {
  527. int result, i, t, pos, idx1, idx2;
  528. IVITile *tile;
  529. int ret = 0;
  530. band->buf = band->bufs[ctx->dst_buf];
  531. band->ref_buf = band->bufs[ctx->ref_buf];
  532. result = decode_band_hdr(ctx, band, avctx);
  533. if (result) {
  534. av_log(avctx, AV_LOG_ERROR, "Error decoding band header\n");
  535. return result;
  536. }
  537. if (band->is_empty) {
  538. av_log(avctx, AV_LOG_ERROR, "Empty band encountered!\n");
  539. return AVERROR_INVALIDDATA;
  540. }
  541. band->rv_map = &ctx->rvmap_tabs[band->rvmap_sel];
  542. /* apply corrections to the selected rvmap table if present */
  543. for (i = 0; i < band->num_corr; i++) {
  544. idx1 = band->corr[i * 2];
  545. idx2 = band->corr[i * 2 + 1];
  546. FFSWAP(uint8_t, band->rv_map->runtab[idx1], band->rv_map->runtab[idx2]);
  547. FFSWAP(int16_t, band->rv_map->valtab[idx1], band->rv_map->valtab[idx2]);
  548. if (idx1 == band->rv_map->eob_sym || idx2 == band->rv_map->eob_sym)
  549. band->rv_map->eob_sym ^= idx1 ^ idx2;
  550. if (idx1 == band->rv_map->esc_sym || idx2 == band->rv_map->esc_sym)
  551. band->rv_map->esc_sym ^= idx1 ^ idx2;
  552. }
  553. pos = get_bits_count(&ctx->gb);
  554. for (t = 0; t < band->num_tiles; t++) {
  555. tile = &band->tiles[t];
  556. tile->is_empty = get_bits1(&ctx->gb);
  557. if (tile->is_empty) {
  558. ff_ivi_process_empty_tile(avctx, band, tile,
  559. (ctx->planes[0].bands[0].mb_size >> 3) - (band->mb_size >> 3));
  560. av_dlog(avctx, "Empty tile encountered!\n");
  561. } else {
  562. tile->data_size = ff_ivi_dec_tile_data_size(&ctx->gb);
  563. if (!tile->data_size) {
  564. av_log(avctx, AV_LOG_ERROR, "Tile data size is zero!\n");
  565. ret = AVERROR_INVALIDDATA;
  566. break;
  567. }
  568. result = decode_mb_info(ctx, band, tile, avctx);
  569. if (result < 0)
  570. break;
  571. result = ff_ivi_decode_blocks(&ctx->gb, band, tile);
  572. if (result < 0 || ((get_bits_count(&ctx->gb) - pos) >> 3) != tile->data_size) {
  573. av_log(avctx, AV_LOG_ERROR, "Corrupted tile data encountered!\n");
  574. break;
  575. }
  576. pos += tile->data_size << 3; // skip to next tile
  577. }
  578. }
  579. /* restore the selected rvmap table by applying its corrections in reverse order */
  580. for (i = band->num_corr - 1; i >= 0; i--) {
  581. idx1 = band->corr[i * 2];
  582. idx2 = band->corr[i * 2 + 1];
  583. FFSWAP(uint8_t, band->rv_map->runtab[idx1], band->rv_map->runtab[idx2]);
  584. FFSWAP(int16_t, band->rv_map->valtab[idx1], band->rv_map->valtab[idx2]);
  585. if (idx1 == band->rv_map->eob_sym || idx2 == band->rv_map->eob_sym)
  586. band->rv_map->eob_sym ^= idx1 ^ idx2;
  587. if (idx1 == band->rv_map->esc_sym || idx2 == band->rv_map->esc_sym)
  588. band->rv_map->esc_sym ^= idx1 ^ idx2;
  589. }
  590. #if defined(DEBUG) && IVI4_DEBUG_CHECKSUM
  591. if (band->checksum_present) {
  592. uint16_t chksum = ivi_calc_band_checksum(band);
  593. if (chksum != band->checksum) {
  594. av_log(avctx, AV_LOG_ERROR,
  595. "Band checksum mismatch! Plane %d, band %d, received: %x, calculated: %x\n",
  596. band->plane, band->band_num, band->checksum, chksum);
  597. }
  598. }
  599. #endif
  600. align_get_bits(&ctx->gb);
  601. return ret;
  602. }
  603. static av_cold int decode_init(AVCodecContext *avctx)
  604. {
  605. IVI4DecContext *ctx = avctx->priv_data;
  606. ff_ivi_init_static_vlc();
  607. /* copy rvmap tables in our context so we can apply changes to them */
  608. 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;
  613. avctx->pix_fmt = PIX_FMT_YUV410P;
  614. return 0;
  615. }
  616. /**
  617. * Rearrange decoding and reference buffers.
  618. *
  619. * @param[in,out] ctx pointer to the decoder context
  620. */
  621. static void switch_buffers(IVI4DecContext *ctx)
  622. {
  623. switch (ctx->prev_frame_type) {
  624. case FRAMETYPE_INTRA:
  625. case FRAMETYPE_INTER:
  626. ctx->buf_switch ^= 1;
  627. ctx->dst_buf = ctx->buf_switch;
  628. ctx->ref_buf = ctx->buf_switch ^ 1;
  629. break;
  630. case FRAMETYPE_INTER_NOREF:
  631. break;
  632. }
  633. switch (ctx->frame_type) {
  634. case FRAMETYPE_INTRA:
  635. ctx->buf_switch = 0;
  636. /* FALLTHROUGH */
  637. case FRAMETYPE_INTER:
  638. ctx->dst_buf = ctx->buf_switch;
  639. ctx->ref_buf = ctx->buf_switch ^ 1;
  640. break;
  641. case FRAMETYPE_INTER_NOREF:
  642. case FRAMETYPE_NULL_FIRST:
  643. case FRAMETYPE_NULL_LAST:
  644. break;
  645. }
  646. }
  647. static int decode_frame(AVCodecContext *avctx, void *data, int *data_size,
  648. AVPacket *avpkt)
  649. {
  650. IVI4DecContext *ctx = avctx->priv_data;
  651. const uint8_t *buf = avpkt->data;
  652. int buf_size = avpkt->size;
  653. int result, p, b;
  654. init_get_bits(&ctx->gb, buf, buf_size * 8);
  655. result = decode_pic_hdr(ctx, avctx);
  656. if (result) {
  657. av_log(avctx, AV_LOG_ERROR, "Error decoding picture header\n");
  658. return result;
  659. }
  660. switch_buffers(ctx);
  661. if (ctx->frame_type < FRAMETYPE_NULL_FIRST) {
  662. for (p = 0; p < 3; p++) {
  663. for (b = 0; b < ctx->planes[p].num_bands; b++) {
  664. result = decode_band(ctx, p, &ctx->planes[p].bands[b], avctx);
  665. if (result) {
  666. av_log(avctx, AV_LOG_ERROR,
  667. "Error decoding band: %d, plane: %d\n", b, p);
  668. return result;
  669. }
  670. }
  671. }
  672. }
  673. /* If the bidirectional mode is enabled, next I and the following P frame will */
  674. /* be sent together. Unfortunately the approach below seems to be the only way */
  675. /* to handle the B-frames mode. That's exactly the same Intel decoders do. */
  676. if (ctx->frame_type == FRAMETYPE_INTRA) {
  677. while (get_bits(&ctx->gb, 8)); // skip version string
  678. skip_bits_long(&ctx->gb, 64); // skip padding, TODO: implement correct 8-bytes alignment
  679. if (get_bits_left(&ctx->gb) > 18 && show_bits(&ctx->gb, 18) == 0x3FFF8)
  680. av_log(avctx, AV_LOG_ERROR, "Buffer contains IP frames!\n");
  681. }
  682. if (ctx->frame_type >= FRAMETYPE_NULL_FIRST)
  683. return buf_size;
  684. if (ctx->frame.data[0])
  685. avctx->release_buffer(avctx, &ctx->frame);
  686. avcodec_set_dimensions(avctx, ctx->planes[0].width, ctx->planes[0].height);
  687. ctx->frame.reference = 0;
  688. if ((result = avctx->get_buffer(avctx, &ctx->frame)) < 0) {
  689. av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
  690. return result;
  691. }
  692. if (ctx->is_scalable) {
  693. ff_ivi_recompose_haar(&ctx->planes[0], ctx->frame.data[0], ctx->frame.linesize[0], 4);
  694. } else {
  695. ff_ivi_output_plane(&ctx->planes[0], ctx->frame.data[0], ctx->frame.linesize[0]);
  696. }
  697. ff_ivi_output_plane(&ctx->planes[2], ctx->frame.data[1], ctx->frame.linesize[1]);
  698. ff_ivi_output_plane(&ctx->planes[1], ctx->frame.data[2], ctx->frame.linesize[2]);
  699. *data_size = sizeof(AVFrame);
  700. *(AVFrame*)data = ctx->frame;
  701. return buf_size;
  702. }
  703. static av_cold int decode_close(AVCodecContext *avctx)
  704. {
  705. IVI4DecContext *ctx = avctx->priv_data;
  706. ff_ivi_free_buffers(&ctx->planes[0]);
  707. if (ctx->frame.data[0])
  708. avctx->release_buffer(avctx, &ctx->frame);
  709. #if IVI4_STREAM_ANALYSER
  710. if (ctx->is_scalable)
  711. av_log(avctx, AV_LOG_ERROR, "This video uses scalability mode!\n");
  712. if (ctx->uses_tiling)
  713. av_log(avctx, AV_LOG_ERROR, "This video uses local decoding!\n");
  714. if (ctx->has_b_frames)
  715. av_log(avctx, AV_LOG_ERROR, "This video contains B-frames!\n");
  716. if (ctx->has_transp)
  717. av_log(avctx, AV_LOG_ERROR, "Transparency mode is enabled!\n");
  718. if (ctx->uses_haar)
  719. av_log(avctx, AV_LOG_ERROR, "This video uses Haar transform!\n");
  720. if (ctx->uses_fullpel)
  721. av_log(avctx, AV_LOG_ERROR, "This video uses fullpel motion vectors!\n");
  722. #endif
  723. return 0;
  724. }
  725. AVCodec ff_indeo4_decoder = {
  726. .name = "indeo4",
  727. .type = AVMEDIA_TYPE_VIDEO,
  728. .id = CODEC_ID_INDEO4,
  729. .priv_data_size = sizeof(IVI4DecContext),
  730. .init = decode_init,
  731. .close = decode_close,
  732. .decode = decode_frame,
  733. .long_name = NULL_IF_CONFIG_SMALL("Intel Indeo Video Interactive 4"),
  734. };