2 * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder
26 #include "libavutil/avassert.h"
28 #include "msmpeg4data.h"
33 typedef struct MSS2Context {
43 static void arith2_normalise(ArithCoder *c)
45 while ((c->high >> 15) - (c->low >> 15) < 2) {
46 if ((c->low ^ c->high) & 0x10000) {
51 c->high = c->high << 8 & 0xFFFFFF | 0xFF;
52 c->value = c->value << 8 & 0xFFFFFF | bytestream2_get_byte(c->gbc.gB);
53 c->low = c->low << 8 & 0xFFFFFF;
59 /* L. Stuiver and A. Moffat: "Piecewise Integer Mapping for Arithmetic Coding."
60 * In Proc. 8th Data Compression Conference (DCC '98), pp. 3-12, Mar. 1998 */
62 static int arith2_get_scaled_value(int value, int n, int range)
64 int split = (n << 1) - range;
67 return split + (value - split >> 1);
72 static void arith2_rescale_interval(ArithCoder *c, int range,
73 int low, int high, int n)
75 int split = (n << 1) - range;
78 c->high = split + (high - split << 1);
82 c->high += c->low - 1;
85 c->low += split + (low - split << 1);
90 static int arith2_get_number(ArithCoder *c, int n)
92 int range = c->high - c->low + 1;
93 int scale = av_log2(range) - av_log2(n);
96 if (n << scale > range)
101 val = arith2_get_scaled_value(c->value - c->low, n, range) >> scale;
103 arith2_rescale_interval(c, range, val << scale, (val + 1) << scale, n);
110 static int arith2_get_prob(ArithCoder *c, int16_t *probs)
112 int range = c->high - c->low + 1, n = *probs;
113 int scale = av_log2(range) - av_log2(n);
116 if (n << scale > range)
121 val = arith2_get_scaled_value(c->value - c->low, n, range) >> scale;
122 while (probs[++i] > val) ;
124 arith2_rescale_interval(c, range,
125 probs[i] << scale, probs[i - 1] << scale, n);
130 ARITH_GET_MODEL_SYM(2)
132 static int arith2_get_consumed_bytes(ArithCoder *c)
134 int diff = (c->high >> 16) - (c->low >> 16);
135 int bp = bytestream2_tell(c->gbc.gB) - 3 << 3;
138 while (!(diff & 0x80)) {
143 return (bits + bp + 7 >> 3) + ((c->low >> 16) + 1 == c->high >> 16);
146 static void arith2_init(ArithCoder *c, GetByteContext *gB)
150 c->value = bytestream2_get_be24(gB);
152 c->get_model_sym = arith2_get_model_sym;
153 c->get_number = arith2_get_number;
156 static int decode_pal_v2(MSS12Context *ctx, const uint8_t *buf, int buf_size)
159 uint32_t *pal = ctx->pal + 256 - ctx->free_colours;
161 if (!ctx->free_colours)
165 if (ncol > ctx->free_colours || buf_size < 2 + ncol * 3)
167 for (i = 0; i < ncol; i++)
168 *pal++ = AV_RB24(buf + 3 * i);
173 static int decode_555(GetByteContext *gB, uint16_t *dst, int stride,
174 int keyframe, int w, int h)
176 int last_symbol = 0, repeat = 0, prev_avail = 0;
179 int x, y, endx, endy, t;
181 #define READ_PAIR(a, b) \
182 a = bytestream2_get_byte(gB) << 4; \
183 t = bytestream2_get_byte(gB); \
185 b = (t & 0xF) << 8; \
186 b |= bytestream2_get_byte(gB); \
191 if (endx >= w || endy >= h || x > endx || y > endy)
193 dst += x + stride * y;
204 int b = bytestream2_get_byte(gB);
206 last_symbol = b << 8 | bytestream2_get_byte(gB);
210 repeat = (repeat << 8) + bytestream2_get_byte(gB) + 1;
211 if (last_symbol == -2) {
212 int skip = FFMIN((unsigned)repeat, dst + w - p);
217 last_symbol = 127 - b;
219 if (last_symbol >= 0)
221 else if (last_symbol == -1 && prev_avail)
223 } while (++p < dst + w);
231 static int decode_rle(GetBitContext *gb, uint8_t *pal_dst, int pal_stride,
232 uint8_t *rgb_dst, int rgb_stride, uint32_t *pal,
233 int keyframe, int kf_slipt, int slice, int w, int h)
235 uint8_t bits[270] = { 0 };
239 int current_length = 0, read_codes = 0, next_code = 0, current_codes = 0;
240 int remaining_codes, surplus_codes, i;
242 const int alphabet_size = 270 - keyframe;
244 int last_symbol = 0, repeat = 0, prev_avail = 0;
247 int x, y, clipw, cliph;
249 x = get_bits(gb, 12);
250 y = get_bits(gb, 12);
251 clipw = get_bits(gb, 12) + 1;
252 cliph = get_bits(gb, 12) + 1;
254 if (x + clipw > w || y + cliph > h)
255 return AVERROR_INVALIDDATA;
256 pal_dst += pal_stride * y + x;
257 rgb_dst += rgb_stride * y + x * 3;
264 pal_dst += pal_stride * kf_slipt;
265 rgb_dst += rgb_stride * kf_slipt;
272 /* read explicit codes */
274 while (current_codes--) {
275 int symbol = get_bits(gb, 8);
276 if (symbol >= 204 - keyframe)
277 symbol += 14 - keyframe;
278 else if (symbol > 189)
279 symbol = get_bits1(gb) + (symbol << 1) - 190;
281 return AVERROR_INVALIDDATA;
282 bits[symbol] = current_length;
283 codes[symbol] = next_code++;
288 remaining_codes = (1 << current_length) - next_code;
289 current_codes = get_bits(gb, av_ceil_log2(remaining_codes + 1));
290 if (current_length > 22 || current_codes > remaining_codes)
291 return AVERROR_INVALIDDATA;
292 } while (current_codes != remaining_codes);
294 remaining_codes = alphabet_size - read_codes;
296 /* determine the minimum length to fit the rest of the alphabet */
297 while ((surplus_codes = (2 << current_length) -
298 (next_code << 1) - remaining_codes) < 0) {
303 /* add the rest of the symbols lexicographically */
304 for (i = 0; i < alphabet_size; i++)
306 if (surplus_codes-- == 0) {
310 bits[i] = current_length;
311 codes[i] = next_code++;
314 if (next_code != 1 << current_length)
315 return AVERROR_INVALIDDATA;
317 if (i = init_vlc(&vlc, 9, alphabet_size, bits, 1, 1, codes, 4, 4, 0))
322 uint8_t *pp = pal_dst;
323 uint8_t *rp = rgb_dst;
326 int b = get_vlc2(gb, vlc.table, 9, 3);
332 b = get_bits(gb, 4) + 10;
337 repeat = get_bits(gb, b);
339 repeat += (1 << b) - 1;
341 if (last_symbol == -2) {
342 int skip = FFMIN(repeat, pal_dst + w - pp);
348 last_symbol = 267 - b;
350 if (last_symbol >= 0) {
352 AV_WB24(rp, pal[last_symbol]);
353 } else if (last_symbol == -1 && prev_avail) {
354 *pp = *(pp - pal_stride);
355 memcpy(rp, rp - rgb_stride, 3);
358 } while (++pp < pal_dst + w);
359 pal_dst += pal_stride;
360 rgb_dst += rgb_stride;
368 static int decode_wmv9(AVCodecContext *avctx, const uint8_t *buf, int buf_size,
369 int x, int y, int w, int h, int wmv9_mask)
371 MSS2Context *ctx = avctx->priv_data;
372 MSS12Context *c = &ctx->c;
373 VC1Context *v = avctx->priv_data;
374 MpegEncContext *s = &v->s;
377 ff_mpeg_flush(avctx);
379 if (s->current_picture_ptr == NULL || s->current_picture_ptr->f.data[0]) {
380 int i = ff_find_unused_picture(s, 0);
383 s->current_picture_ptr = &s->picture[i];
386 init_get_bits(&s->gb, buf, buf_size * 8);
388 s->loop_filter = avctx->skip_loop_filter < AVDISCARD_ALL;
390 if (ff_vc1_parse_frame_header(v, &s->gb) == -1) {
391 av_log(v->s.avctx, AV_LOG_ERROR, "header error\n");
392 return AVERROR_INVALIDDATA;
395 if (s->pict_type != AV_PICTURE_TYPE_I) {
396 av_log(v->s.avctx, AV_LOG_ERROR, "expected I-frame\n");
397 return AVERROR_INVALIDDATA;
400 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
402 if (ff_MPV_frame_start(s, avctx) < 0) {
403 av_log(v->s.avctx, AV_LOG_ERROR, "ff_MPV_frame_start error\n");
404 avctx->pix_fmt = AV_PIX_FMT_RGB24;
408 ff_er_frame_start(s);
410 v->bits = buf_size * 8;
412 v->end_mb_x = (w + 15) >> 4;
413 s->end_mb_y = (h + 15) >> 4;
415 v->end_mb_x = v->end_mb_x + 1 >> 1;
417 s->end_mb_y = s->end_mb_y + 1 >> 1;
419 ff_vc1_decode_blocks(v);
425 f = &s->current_picture.f;
427 if (v->respic == 3) {
428 ctx->dsp.upsample_plane(f->data[0], f->linesize[0], w, h);
429 ctx->dsp.upsample_plane(f->data[1], f->linesize[1], w >> 1, h >> 1);
430 ctx->dsp.upsample_plane(f->data[2], f->linesize[2], w >> 1, h >> 1);
431 } else if (v->respic)
432 av_log_ask_for_sample(v->s.avctx,
433 "Asymmetric WMV9 rectangle subsampling\n");
435 av_assert0(f->linesize[1] == f->linesize[2]);
438 ctx->dsp.mss2_blit_wmv9_masked(c->rgb_pic + y * c->rgb_stride + x * 3,
439 c->rgb_stride, wmv9_mask,
440 c->pal_pic + y * c->pal_stride + x,
442 f->data[0], f->linesize[0],
443 f->data[1], f->data[2], f->linesize[1],
446 ctx->dsp.mss2_blit_wmv9(c->rgb_pic + y * c->rgb_stride + x * 3,
448 f->data[0], f->linesize[0],
449 f->data[1], f->data[2], f->linesize[1],
452 avctx->pix_fmt = AV_PIX_FMT_RGB24;
457 typedef struct Rectangle {
458 int coded, x, y, w, h;
461 #define MAX_WMV9_RECTANGLES 20
462 #define ARITH2_PADDING 2
464 static int mss2_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
467 const uint8_t *buf = avpkt->data;
468 int buf_size = avpkt->size;
469 MSS2Context *ctx = avctx->priv_data;
470 MSS12Context *c = &ctx->c;
475 int keyframe, has_wmv9, has_mv, is_rle, is_555, ret;
477 Rectangle wmv9rects[MAX_WMV9_RECTANGLES], *r;
478 int used_rects = 0, i, implicit_rect = 0, av_uninit(wmv9_mask);
480 av_assert0(FF_INPUT_BUFFER_PADDING_SIZE >=
481 ARITH2_PADDING + (MIN_CACHE_BITS + 7) / 8);
483 init_get_bits(&gb, buf, buf_size * 8);
485 if (keyframe = get_bits1(&gb))
487 has_wmv9 = get_bits1(&gb);
488 has_mv = keyframe ? 0 : get_bits1(&gb);
489 is_rle = get_bits1(&gb);
490 is_555 = is_rle && get_bits1(&gb);
491 if (c->slice_split > 0)
492 ctx->split_position = c->slice_split;
493 else if (c->slice_split < 0) {
494 if (get_bits1(&gb)) {
495 if (get_bits1(&gb)) {
497 ctx->split_position = get_bits(&gb, 16);
499 ctx->split_position = get_bits(&gb, 12);
501 ctx->split_position = get_bits(&gb, 8) << 4;
504 ctx->split_position = avctx->height / 2;
507 ctx->split_position = avctx->height;
509 if (c->slice_split && (ctx->split_position < 1 - is_555 ||
510 ctx->split_position > avctx->height - 1))
511 return AVERROR_INVALIDDATA;
514 buf += get_bits_count(&gb) >> 3;
515 buf_size -= get_bits_count(&gb) >> 3;
518 return AVERROR_INVALIDDATA;
520 if (is_555 && (has_wmv9 || has_mv || c->slice_split && ctx->split_position))
521 return AVERROR_INVALIDDATA;
523 avctx->pix_fmt = is_555 ? AV_PIX_FMT_RGB555 : AV_PIX_FMT_RGB24;
524 if (ctx->pic.data[0] && ctx->pic.format != avctx->pix_fmt)
525 avctx->release_buffer(avctx, &ctx->pic);
528 bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
529 arith2_init(&acoder, &gB);
531 implicit_rect = !arith2_get_bit(&acoder);
533 while (arith2_get_bit(&acoder)) {
534 if (used_rects == MAX_WMV9_RECTANGLES)
535 return AVERROR_INVALIDDATA;
536 r = &wmv9rects[used_rects];
538 r->x = arith2_get_number(&acoder, avctx->width);
540 r->x = arith2_get_number(&acoder, avctx->width -
541 wmv9rects[used_rects - 1].x) +
542 wmv9rects[used_rects - 1].x;
543 r->y = arith2_get_number(&acoder, avctx->height);
544 r->w = arith2_get_number(&acoder, avctx->width - r->x) + 1;
545 r->h = arith2_get_number(&acoder, avctx->height - r->y) + 1;
549 if (implicit_rect && used_rects) {
550 av_log(avctx, AV_LOG_ERROR, "implicit_rect && used_rects > 0\n");
551 return AVERROR_INVALIDDATA;
557 wmv9rects[0].w = avctx->width;
558 wmv9rects[0].h = avctx->height;
562 for (i = 0; i < used_rects; i++) {
563 if (!implicit_rect && arith2_get_bit(&acoder)) {
564 av_log(avctx, AV_LOG_ERROR, "Unexpected grandchildren\n");
565 return AVERROR_INVALIDDATA;
568 wmv9_mask = arith2_get_bit(&acoder) - 1;
570 wmv9_mask = arith2_get_number(&acoder, 256);
572 wmv9rects[i].coded = arith2_get_number(&acoder, 2);
575 buf += arith2_get_consumed_bytes(&acoder);
576 buf_size -= arith2_get_consumed_bytes(&acoder);
578 return AVERROR_INVALIDDATA;
582 if (keyframe && !is_555) {
583 if ((i = decode_pal_v2(c, buf, buf_size)) < 0)
584 return AVERROR_INVALIDDATA;
591 return AVERROR_INVALIDDATA;
592 c->mvX = AV_RB16(buf - 4) - avctx->width;
593 c->mvY = AV_RB16(buf - 2) - avctx->height;
596 if (c->mvX < 0 || c->mvY < 0) {
597 FFSWAP(AVFrame, ctx->pic, ctx->last_pic);
598 FFSWAP(uint8_t *, c->pal_pic, c->last_pal_pic);
600 if (ctx->pic.data[0])
601 avctx->release_buffer(avctx, &ctx->pic);
603 ctx->pic.reference = 3;
604 ctx->pic.buffer_hints = FF_BUFFER_HINTS_VALID |
605 FF_BUFFER_HINTS_READABLE |
606 FF_BUFFER_HINTS_PRESERVE |
607 FF_BUFFER_HINTS_REUSABLE;
609 if ((ret = ff_get_buffer(avctx, &ctx->pic)) < 0) {
610 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
614 if (ctx->last_pic.data[0]) {
615 av_assert0(ctx->pic.linesize[0] == ctx->last_pic.linesize[0]);
616 c->last_rgb_pic = ctx->last_pic.data[0] +
617 ctx->last_pic.linesize[0] * (avctx->height - 1);
619 av_log(avctx, AV_LOG_ERROR, "Missing keyframe\n");
623 if (ctx->last_pic.data[0])
624 avctx->release_buffer(avctx, &ctx->last_pic);
626 ctx->pic.reference = 3;
627 ctx->pic.buffer_hints = FF_BUFFER_HINTS_VALID |
628 FF_BUFFER_HINTS_READABLE |
629 FF_BUFFER_HINTS_PRESERVE |
630 FF_BUFFER_HINTS_REUSABLE;
632 if ((ret = avctx->reget_buffer(avctx, &ctx->pic)) < 0) {
633 av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n");
637 c->last_rgb_pic = NULL;
639 c->rgb_pic = ctx->pic.data[0] +
640 ctx->pic.linesize[0] * (avctx->height - 1);
641 c->rgb_stride = -ctx->pic.linesize[0];
643 ctx->pic.key_frame = keyframe;
644 ctx->pic.pict_type = keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
647 bytestream2_init(&gB, buf, buf_size);
649 if (decode_555(&gB, (uint16_t *)c->rgb_pic, c->rgb_stride >> 1,
650 keyframe, avctx->width, avctx->height))
651 return AVERROR_INVALIDDATA;
653 buf_size -= bytestream2_tell(&gB);
657 ff_mss12_slicecontext_reset(&ctx->sc[0]);
659 ff_mss12_slicecontext_reset(&ctx->sc[1]);
662 init_get_bits(&gb, buf, buf_size * 8);
663 if (ret = decode_rle(&gb, c->pal_pic, c->pal_stride,
664 c->rgb_pic, c->rgb_stride, c->pal, keyframe,
665 ctx->split_position, 0,
666 avctx->width, avctx->height))
671 if (ret = decode_rle(&gb, c->pal_pic, c->pal_stride,
672 c->rgb_pic, c->rgb_stride, c->pal, keyframe,
673 ctx->split_position, 1,
674 avctx->width, avctx->height))
678 buf += get_bits_count(&gb) >> 3;
679 buf_size -= get_bits_count(&gb) >> 3;
680 } else if (!implicit_rect || wmv9_mask != -1) {
682 return AVERROR_INVALIDDATA;
683 bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
684 arith2_init(&acoder, &gB);
685 c->keyframe = keyframe;
686 if (c->corrupted = ff_mss12_decode_rect(&ctx->sc[0], &acoder, 0, 0,
688 ctx->split_position))
689 return AVERROR_INVALIDDATA;
691 buf += arith2_get_consumed_bytes(&acoder);
692 buf_size -= arith2_get_consumed_bytes(&acoder);
693 if (c->slice_split) {
695 return AVERROR_INVALIDDATA;
696 bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
697 arith2_init(&acoder, &gB);
698 if (c->corrupted = ff_mss12_decode_rect(&ctx->sc[1], &acoder, 0,
701 avctx->height - ctx->split_position))
702 return AVERROR_INVALIDDATA;
704 buf += arith2_get_consumed_bytes(&acoder);
705 buf_size -= arith2_get_consumed_bytes(&acoder);
708 memset(c->pal_pic, 0, c->pal_stride * avctx->height);
712 for (i = 0; i < used_rects; i++) {
713 int x = wmv9rects[i].x;
714 int y = wmv9rects[i].y;
715 int w = wmv9rects[i].w;
716 int h = wmv9rects[i].h;
717 if (wmv9rects[i].coded) {
718 int WMV9codedFrameSize;
719 if (buf_size < 4 || !(WMV9codedFrameSize = AV_RL24(buf)))
720 return AVERROR_INVALIDDATA;
721 if (ret = decode_wmv9(avctx, buf + 3, buf_size - 3,
722 x, y, w, h, wmv9_mask))
724 buf += WMV9codedFrameSize + 3;
725 buf_size -= WMV9codedFrameSize + 3;
727 uint8_t *dst = c->rgb_pic + y * c->rgb_stride + x * 3;
728 if (wmv9_mask != -1) {
729 ctx->dsp.mss2_gray_fill_masked(dst, c->rgb_stride,
731 c->pal_pic + y * c->pal_stride + x,
736 memset(dst, 0x80, w * 3);
737 dst += c->rgb_stride;
745 av_log(avctx, AV_LOG_WARNING, "buffer not fully consumed\n");
748 *(AVFrame *)data = ctx->pic;
753 static av_cold int wmv9_init(AVCodecContext *avctx)
755 VC1Context *v = avctx->priv_data;
758 avctx->flags |= CODEC_FLAG_EMU_EDGE;
759 v->s.flags |= CODEC_FLAG_EMU_EDGE;
761 if (avctx->idct_algo == FF_IDCT_AUTO)
762 avctx->idct_algo = FF_IDCT_WMV2;
764 if (ff_vc1_init_common(v) < 0)
766 ff_vc1dsp_init(&v->vc1dsp);
768 v->profile = PROFILE_MAIN;
770 v->zz_8x4 = ff_wmv2_scantableA;
771 v->zz_4x8 = ff_wmv2_scantableB;
775 v->frmrtq_postproc = 7;
776 v->bitrtq_postproc = 31;
793 v->s.resync_marker = 0;
796 v->s.max_b_frames = avctx->max_b_frames = 0;
797 v->quantizer_mode = 0;
803 ff_vc1_init_transposed_scantables(v);
805 if (ff_msmpeg4_decode_init(avctx) < 0 ||
806 ff_vc1_decode_init_alloc_tables(v) < 0)
809 /* error concealment */
810 v->s.me.qpel_put = v->s.dsp.put_qpel_pixels_tab;
811 v->s.me.qpel_avg = v->s.dsp.avg_qpel_pixels_tab;
816 static av_cold int mss2_decode_end(AVCodecContext *avctx)
818 MSS2Context *const ctx = avctx->priv_data;
820 if (ctx->pic.data[0])
821 avctx->release_buffer(avctx, &ctx->pic);
822 if (ctx->last_pic.data[0])
823 avctx->release_buffer(avctx, &ctx->last_pic);
825 ff_mss12_decode_end(&ctx->c);
826 av_freep(&ctx->c.pal_pic);
827 av_freep(&ctx->c.last_pal_pic);
828 ff_vc1_decode_end(avctx);
833 static av_cold int mss2_decode_init(AVCodecContext *avctx)
835 MSS2Context * const ctx = avctx->priv_data;
836 MSS12Context *c = &ctx->c;
839 avctx->coded_frame = &ctx->pic;
840 if (ret = ff_mss12_decode_init(c, 1, &ctx->sc[0], &ctx->sc[1]))
842 c->pal_stride = c->mask_stride;
843 c->pal_pic = av_mallocz(c->pal_stride * avctx->height);
844 c->last_pal_pic = av_mallocz(c->pal_stride * avctx->height);
845 if (!c->pal_pic || !c->last_pal_pic) {
846 mss2_decode_end(avctx);
847 return AVERROR(ENOMEM);
849 if (ret = wmv9_init(avctx)) {
850 mss2_decode_end(avctx);
853 ff_mss2dsp_init(&ctx->dsp);
855 avctx->pix_fmt = c->free_colours == 127 ? AV_PIX_FMT_RGB555
861 AVCodec ff_mss2_decoder = {
863 .type = AVMEDIA_TYPE_VIDEO,
864 .id = AV_CODEC_ID_MSS2,
865 .priv_data_size = sizeof(MSS2Context),
866 .init = mss2_decode_init,
867 .close = mss2_decode_end,
868 .decode = mss2_decode_frame,
869 .capabilities = CODEC_CAP_DR1,
870 .long_name = NULL_IF_CONFIG_SMALL("MS Windows Media Video V9 Screen"),