2 * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder
4 * This file is part of Libav.
6 * Libav 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 * Libav 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 Libav; 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"
27 #include "error_resilience.h"
30 #include "msmpeg4data.h"
36 typedef struct MSS2Context {
46 static void arith2_normalise(ArithCoder *c)
48 while ((c->high >> 15) - (c->low >> 15) < 2) {
49 if ((c->low ^ c->high) & 0x10000) {
54 c->high = c->high << 8 & 0xFFFFFF | 0xFF;
55 c->value = c->value << 8 & 0xFFFFFF | bytestream2_get_byte(c->gbc.gB);
56 c->low = c->low << 8 & 0xFFFFFF;
62 /* L. Stuiver and A. Moffat: "Piecewise Integer Mapping for Arithmetic Coding."
63 * In Proc. 8th Data Compression Conference (DCC '98), pp. 3-12, Mar. 1998 */
65 static int arith2_get_scaled_value(int value, int n, int range)
67 int split = (n << 1) - range;
70 return split + (value - split >> 1);
75 static void arith2_rescale_interval(ArithCoder *c, int range,
76 int low, int high, int n)
78 int split = (n << 1) - range;
81 c->high = split + (high - split << 1);
85 c->high += c->low - 1;
88 c->low += split + (low - split << 1);
93 static int arith2_get_number(ArithCoder *c, int n)
95 int range = c->high - c->low + 1;
96 int scale = av_log2(range) - av_log2(n);
99 if (n << scale > range)
104 val = arith2_get_scaled_value(c->value - c->low, n, range) >> scale;
106 arith2_rescale_interval(c, range, val << scale, (val + 1) << scale, n);
113 static int arith2_get_prob(ArithCoder *c, int16_t *probs)
115 int range = c->high - c->low + 1, n = *probs;
116 int scale = av_log2(range) - av_log2(n);
119 if (n << scale > range)
124 val = arith2_get_scaled_value(c->value - c->low, n, range) >> scale;
125 while (probs[++i] > val) ;
127 arith2_rescale_interval(c, range,
128 probs[i] << scale, probs[i - 1] << scale, n);
133 ARITH_GET_MODEL_SYM(2)
135 static int arith2_get_consumed_bytes(ArithCoder *c)
137 int diff = (c->high >> 16) - (c->low >> 16);
138 int bp = bytestream2_tell(c->gbc.gB) - 3 << 3;
141 while (!(diff & 0x80)) {
146 return (bits + bp + 7 >> 3) + ((c->low >> 16) + 1 == c->high >> 16);
149 static void arith2_init(ArithCoder *c, GetByteContext *gB)
153 c->value = bytestream2_get_be24(gB);
155 c->get_model_sym = arith2_get_model_sym;
156 c->get_number = arith2_get_number;
159 static int decode_pal_v2(MSS12Context *ctx, const uint8_t *buf, int buf_size)
162 uint32_t *pal = ctx->pal + 256 - ctx->free_colours;
164 if (!ctx->free_colours)
168 if (ncol > ctx->free_colours || buf_size < 2 + ncol * 3)
169 return AVERROR_INVALIDDATA;
170 for (i = 0; i < ncol; i++)
171 *pal++ = AV_RB24(buf + 3 * i);
176 static int decode_555(GetByteContext *gB, uint16_t *dst, int stride,
177 int keyframe, int w, int h)
179 int last_symbol = 0, repeat = 0, prev_avail = 0;
182 int x, y, endx, endy, t;
184 #define READ_PAIR(a, b) \
185 a = bytestream2_get_byte(gB) << 4; \
186 t = bytestream2_get_byte(gB); \
188 b = (t & 0xF) << 8; \
189 b |= bytestream2_get_byte(gB); \
194 if (endx >= w || endy >= h || x > endx || y > endy)
195 return AVERROR_INVALIDDATA;
196 dst += x + stride * y;
207 int b = bytestream2_get_byte(gB);
209 last_symbol = b << 8 | bytestream2_get_byte(gB);
213 repeat = (repeat << 8) + bytestream2_get_byte(gB) + 1;
214 if (last_symbol == -2) {
215 int skip = FFMIN((unsigned)repeat, dst + w - p);
220 last_symbol = 127 - b;
222 if (last_symbol >= 0)
224 else if (last_symbol == -1 && prev_avail)
226 } while (++p < dst + w);
234 static int decode_rle(GetBitContext *gb, uint8_t *pal_dst, int pal_stride,
235 uint8_t *rgb_dst, int rgb_stride, uint32_t *pal,
236 int keyframe, int kf_slipt, int slice, int w, int h)
238 uint8_t bits[270] = { 0 };
242 int current_length = 0, read_codes = 0, next_code = 0, current_codes = 0;
243 int remaining_codes, surplus_codes, i;
245 const int alphabet_size = 270 - keyframe;
247 int last_symbol = 0, repeat = 0, prev_avail = 0;
250 int x, y, clipw, cliph;
252 x = get_bits(gb, 12);
253 y = get_bits(gb, 12);
254 clipw = get_bits(gb, 12) + 1;
255 cliph = get_bits(gb, 12) + 1;
257 if (x + clipw > w || y + cliph > h)
258 return AVERROR_INVALIDDATA;
259 pal_dst += pal_stride * y + x;
260 rgb_dst += rgb_stride * y + x * 3;
267 pal_dst += pal_stride * kf_slipt;
268 rgb_dst += rgb_stride * kf_slipt;
275 /* read explicit codes */
277 while (current_codes--) {
278 int symbol = get_bits(gb, 8);
279 if (symbol >= 204 - keyframe)
280 symbol += 14 - keyframe;
281 else if (symbol > 189)
282 symbol = get_bits1(gb) + (symbol << 1) - 190;
284 return AVERROR_INVALIDDATA;
285 bits[symbol] = current_length;
286 codes[symbol] = next_code++;
291 remaining_codes = (1 << current_length) - next_code;
292 current_codes = get_bits(gb, av_ceil_log2(remaining_codes + 1));
293 if (current_length > 22 || current_codes > remaining_codes)
294 return AVERROR_INVALIDDATA;
295 } while (current_codes != remaining_codes);
297 remaining_codes = alphabet_size - read_codes;
299 /* determine the minimum length to fit the rest of the alphabet */
300 while ((surplus_codes = (2 << current_length) -
301 (next_code << 1) - remaining_codes) < 0) {
306 /* add the rest of the symbols lexicographically */
307 for (i = 0; i < alphabet_size; i++)
309 if (surplus_codes-- == 0) {
313 bits[i] = current_length;
314 codes[i] = next_code++;
317 if (next_code != 1 << current_length)
318 return AVERROR_INVALIDDATA;
320 if (i = init_vlc(&vlc, 9, alphabet_size, bits, 1, 1, codes, 4, 4, 0))
325 uint8_t *pp = pal_dst;
326 uint8_t *rp = rgb_dst;
329 int b = get_vlc2(gb, vlc.table, 9, 3);
335 b = get_bits(gb, 4) + 10;
340 repeat = get_bits(gb, b);
342 repeat += (1 << b) - 1;
344 if (last_symbol == -2) {
345 int skip = FFMIN(repeat, pal_dst + w - pp);
351 last_symbol = 267 - b;
353 if (last_symbol >= 0) {
355 AV_WB24(rp, pal[last_symbol]);
356 } else if (last_symbol == -1 && prev_avail) {
357 *pp = *(pp - pal_stride);
358 memcpy(rp, rp - rgb_stride, 3);
361 } while (++pp < pal_dst + w);
362 pal_dst += pal_stride;
363 rgb_dst += rgb_stride;
371 static int decode_wmv9(AVCodecContext *avctx, const uint8_t *buf, int buf_size,
372 int x, int y, int w, int h, int wmv9_mask)
374 MSS2Context *ctx = avctx->priv_data;
375 MSS12Context *c = &ctx->c;
376 VC1Context *v = avctx->priv_data;
377 MpegEncContext *s = &v->s;
381 ff_mpeg_flush(avctx);
383 init_get_bits(&s->gb, buf, buf_size * 8);
385 s->loop_filter = avctx->skip_loop_filter < AVDISCARD_ALL;
387 if (ff_vc1_parse_frame_header(v, &s->gb) < 0) {
388 av_log(v->s.avctx, AV_LOG_ERROR, "header error\n");
389 return AVERROR_INVALIDDATA;
392 if (s->pict_type != AV_PICTURE_TYPE_I) {
393 av_log(v->s.avctx, AV_LOG_ERROR, "expected I-frame\n");
394 return AVERROR_INVALIDDATA;
397 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
399 if ((ret = ff_MPV_frame_start(s, avctx)) < 0) {
400 av_log(v->s.avctx, AV_LOG_ERROR, "ff_MPV_frame_start error\n");
401 avctx->pix_fmt = AV_PIX_FMT_RGB24;
405 ff_mpeg_er_frame_start(s);
407 v->bits = buf_size * 8;
409 v->end_mb_x = (w + 15) >> 4;
410 s->end_mb_y = (h + 15) >> 4;
412 v->end_mb_x = v->end_mb_x + 1 >> 1;
414 s->end_mb_y = s->end_mb_y + 1 >> 1;
416 ff_vc1_decode_blocks(v);
418 ff_er_frame_end(&s->er);
422 f = s->current_picture.f;
424 if (v->respic == 3) {
425 ctx->dsp.upsample_plane(f->data[0], f->linesize[0], w, h);
426 ctx->dsp.upsample_plane(f->data[1], f->linesize[1], w >> 1, h >> 1);
427 ctx->dsp.upsample_plane(f->data[2], f->linesize[2], w >> 1, h >> 1);
428 } else if (v->respic)
429 avpriv_request_sample(v->s.avctx,
430 "Asymmetric WMV9 rectangle subsampling");
432 av_assert0(f->linesize[1] == f->linesize[2]);
435 ctx->dsp.mss2_blit_wmv9_masked(c->rgb_pic + y * c->rgb_stride + x * 3,
436 c->rgb_stride, wmv9_mask,
437 c->pal_pic + y * c->pal_stride + x,
439 f->data[0], f->linesize[0],
440 f->data[1], f->data[2], f->linesize[1],
443 ctx->dsp.mss2_blit_wmv9(c->rgb_pic + y * c->rgb_stride + x * 3,
445 f->data[0], f->linesize[0],
446 f->data[1], f->data[2], f->linesize[1],
449 avctx->pix_fmt = AV_PIX_FMT_RGB24;
454 typedef struct Rectangle {
455 int coded, x, y, w, h;
458 #define MAX_WMV9_RECTANGLES 20
459 #define ARITH2_PADDING 2
461 static int mss2_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
464 const uint8_t *buf = avpkt->data;
465 int buf_size = avpkt->size;
466 MSS2Context *ctx = avctx->priv_data;
467 MSS12Context *c = &ctx->c;
468 AVFrame *frame = data;
473 int keyframe, has_wmv9, has_mv, is_rle, is_555, ret;
475 Rectangle wmv9rects[MAX_WMV9_RECTANGLES], *r;
476 int used_rects = 0, i, implicit_rect = 0, av_uninit(wmv9_mask);
478 av_assert0(FF_INPUT_BUFFER_PADDING_SIZE >=
479 ARITH2_PADDING + (MIN_CACHE_BITS + 7) / 8);
481 init_get_bits(&gb, buf, buf_size * 8);
483 if (keyframe = get_bits1(&gb))
485 has_wmv9 = get_bits1(&gb);
486 has_mv = keyframe ? 0 : get_bits1(&gb);
487 is_rle = get_bits1(&gb);
488 is_555 = is_rle && get_bits1(&gb);
489 if (c->slice_split > 0)
490 ctx->split_position = c->slice_split;
491 else if (c->slice_split < 0) {
492 if (get_bits1(&gb)) {
493 if (get_bits1(&gb)) {
495 ctx->split_position = get_bits(&gb, 16);
497 ctx->split_position = get_bits(&gb, 12);
499 ctx->split_position = get_bits(&gb, 8) << 4;
502 ctx->split_position = avctx->height / 2;
505 ctx->split_position = avctx->height;
507 if (c->slice_split && (ctx->split_position < 1 - is_555 ||
508 ctx->split_position > avctx->height - 1))
509 return AVERROR_INVALIDDATA;
512 buf += get_bits_count(&gb) >> 3;
513 buf_size -= get_bits_count(&gb) >> 3;
516 return AVERROR_INVALIDDATA;
518 if (is_555 && (has_wmv9 || has_mv || c->slice_split && ctx->split_position))
519 return AVERROR_INVALIDDATA;
521 avctx->pix_fmt = is_555 ? AV_PIX_FMT_RGB555 : AV_PIX_FMT_RGB24;
522 if (ctx->last_pic->format != avctx->pix_fmt)
523 av_frame_unref(ctx->last_pic);
526 bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
527 arith2_init(&acoder, &gB);
529 implicit_rect = !arith2_get_bit(&acoder);
531 while (arith2_get_bit(&acoder)) {
532 if (used_rects == MAX_WMV9_RECTANGLES)
533 return AVERROR_INVALIDDATA;
534 r = &wmv9rects[used_rects];
536 r->x = arith2_get_number(&acoder, avctx->width);
538 r->x = arith2_get_number(&acoder, avctx->width -
539 wmv9rects[used_rects - 1].x) +
540 wmv9rects[used_rects - 1].x;
541 r->y = arith2_get_number(&acoder, avctx->height);
542 r->w = arith2_get_number(&acoder, avctx->width - r->x) + 1;
543 r->h = arith2_get_number(&acoder, avctx->height - r->y) + 1;
547 if (implicit_rect && used_rects) {
548 av_log(avctx, AV_LOG_ERROR, "implicit_rect && used_rects > 0\n");
549 return AVERROR_INVALIDDATA;
555 wmv9rects[0].w = avctx->width;
556 wmv9rects[0].h = avctx->height;
560 for (i = 0; i < used_rects; i++) {
561 if (!implicit_rect && arith2_get_bit(&acoder)) {
562 av_log(avctx, AV_LOG_ERROR, "Unexpected grandchildren\n");
563 return AVERROR_INVALIDDATA;
566 wmv9_mask = arith2_get_bit(&acoder) - 1;
568 wmv9_mask = arith2_get_number(&acoder, 256);
570 wmv9rects[i].coded = arith2_get_number(&acoder, 2);
573 buf += arith2_get_consumed_bytes(&acoder);
574 buf_size -= arith2_get_consumed_bytes(&acoder);
576 return AVERROR_INVALIDDATA;
580 if (keyframe && !is_555) {
581 if ((i = decode_pal_v2(c, buf, buf_size)) < 0)
582 return AVERROR_INVALIDDATA;
589 return AVERROR_INVALIDDATA;
590 c->mvX = AV_RB16(buf - 4) - avctx->width;
591 c->mvY = AV_RB16(buf - 2) - avctx->height;
594 if (c->mvX < 0 || c->mvY < 0) {
595 FFSWAP(uint8_t *, c->pal_pic, c->last_pal_pic);
597 if ((ret = ff_get_buffer(avctx, frame, AV_GET_BUFFER_FLAG_REF)) < 0) {
598 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
602 if (ctx->last_pic->data[0]) {
603 av_assert0(frame->linesize[0] == ctx->last_pic->linesize[0]);
604 c->last_rgb_pic = ctx->last_pic->data[0] +
605 ctx->last_pic->linesize[0] * (avctx->height - 1);
607 av_log(avctx, AV_LOG_ERROR, "Missing keyframe\n");
608 return AVERROR_INVALIDDATA;
611 if ((ret = ff_reget_buffer(avctx, ctx->last_pic)) < 0) {
612 av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n");
615 if ((ret = av_frame_ref(frame, ctx->last_pic)) < 0)
618 c->last_rgb_pic = NULL;
620 c->rgb_pic = frame->data[0] +
621 frame->linesize[0] * (avctx->height - 1);
622 c->rgb_stride = -frame->linesize[0];
624 frame->key_frame = keyframe;
625 frame->pict_type = keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
628 bytestream2_init(&gB, buf, buf_size);
630 if (decode_555(&gB, (uint16_t *)c->rgb_pic, c->rgb_stride >> 1,
631 keyframe, avctx->width, avctx->height))
632 return AVERROR_INVALIDDATA;
634 buf_size -= bytestream2_tell(&gB);
638 ff_mss12_slicecontext_reset(&ctx->sc[0]);
640 ff_mss12_slicecontext_reset(&ctx->sc[1]);
643 init_get_bits(&gb, buf, buf_size * 8);
644 if (ret = decode_rle(&gb, c->pal_pic, c->pal_stride,
645 c->rgb_pic, c->rgb_stride, c->pal, keyframe,
646 ctx->split_position, 0,
647 avctx->width, avctx->height))
652 if (ret = decode_rle(&gb, c->pal_pic, c->pal_stride,
653 c->rgb_pic, c->rgb_stride, c->pal, keyframe,
654 ctx->split_position, 1,
655 avctx->width, avctx->height))
659 buf += get_bits_count(&gb) >> 3;
660 buf_size -= get_bits_count(&gb) >> 3;
661 } else if (!implicit_rect || wmv9_mask != -1) {
663 return AVERROR_INVALIDDATA;
664 bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
665 arith2_init(&acoder, &gB);
666 c->keyframe = keyframe;
667 if (c->corrupted = ff_mss12_decode_rect(&ctx->sc[0], &acoder, 0, 0,
669 ctx->split_position))
670 return AVERROR_INVALIDDATA;
672 buf += arith2_get_consumed_bytes(&acoder);
673 buf_size -= arith2_get_consumed_bytes(&acoder);
674 if (c->slice_split) {
676 return AVERROR_INVALIDDATA;
677 bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
678 arith2_init(&acoder, &gB);
679 if (c->corrupted = ff_mss12_decode_rect(&ctx->sc[1], &acoder, 0,
682 avctx->height - ctx->split_position))
683 return AVERROR_INVALIDDATA;
685 buf += arith2_get_consumed_bytes(&acoder);
686 buf_size -= arith2_get_consumed_bytes(&acoder);
689 memset(c->pal_pic, 0, c->pal_stride * avctx->height);
693 for (i = 0; i < used_rects; i++) {
694 int x = wmv9rects[i].x;
695 int y = wmv9rects[i].y;
696 int w = wmv9rects[i].w;
697 int h = wmv9rects[i].h;
698 if (wmv9rects[i].coded) {
699 int WMV9codedFrameSize;
700 if (buf_size < 4 || !(WMV9codedFrameSize = AV_RL24(buf)))
701 return AVERROR_INVALIDDATA;
702 if (ret = decode_wmv9(avctx, buf + 3, buf_size - 3,
703 x, y, w, h, wmv9_mask))
705 buf += WMV9codedFrameSize + 3;
706 buf_size -= WMV9codedFrameSize + 3;
708 uint8_t *dst = c->rgb_pic + y * c->rgb_stride + x * 3;
709 if (wmv9_mask != -1) {
710 ctx->dsp.mss2_gray_fill_masked(dst, c->rgb_stride,
712 c->pal_pic + y * c->pal_stride + x,
717 memset(dst, 0x80, w * 3);
718 dst += c->rgb_stride;
726 av_log(avctx, AV_LOG_WARNING, "buffer not fully consumed\n");
728 if (c->mvX < 0 || c->mvY < 0) {
729 av_frame_unref(ctx->last_pic);
730 ret = av_frame_ref(ctx->last_pic, frame);
740 static av_cold int wmv9_init(AVCodecContext *avctx)
742 VC1Context *v = avctx->priv_data;
747 if ((ret = ff_vc1_init_common(v)) < 0)
749 ff_vc1dsp_init(&v->vc1dsp);
751 v->profile = PROFILE_MAIN;
753 v->zz_8x4 = ff_wmv2_scantableA;
754 v->zz_4x8 = ff_wmv2_scantableB;
758 v->frmrtq_postproc = 7;
759 v->bitrtq_postproc = 31;
776 v->resync_marker = 0;
779 v->s.max_b_frames = avctx->max_b_frames = 0;
780 v->quantizer_mode = 0;
786 ff_vc1_init_transposed_scantables(v);
788 if ((ret = ff_msmpeg4_decode_init(avctx)) < 0 ||
789 (ret = ff_vc1_decode_init_alloc_tables(v)) < 0)
792 /* error concealment */
793 v->s.me.qpel_put = v->s.qdsp.put_qpel_pixels_tab;
794 v->s.me.qpel_avg = v->s.qdsp.avg_qpel_pixels_tab;
799 static av_cold int mss2_decode_end(AVCodecContext *avctx)
801 MSS2Context *const ctx = avctx->priv_data;
803 av_frame_free(&ctx->last_pic);
805 ff_mss12_decode_end(&ctx->c);
806 av_freep(&ctx->c.pal_pic);
807 av_freep(&ctx->c.last_pal_pic);
808 ff_vc1_decode_end(avctx);
813 static av_cold int mss2_decode_init(AVCodecContext *avctx)
815 MSS2Context * const ctx = avctx->priv_data;
816 MSS12Context *c = &ctx->c;
819 if (ret = ff_mss12_decode_init(c, 1, &ctx->sc[0], &ctx->sc[1]))
821 c->pal_stride = c->mask_stride;
822 c->pal_pic = av_mallocz(c->pal_stride * avctx->height);
823 c->last_pal_pic = av_mallocz(c->pal_stride * avctx->height);
824 if (!c->pal_pic || !c->last_pal_pic) {
825 mss2_decode_end(avctx);
826 return AVERROR(ENOMEM);
828 if (ret = wmv9_init(avctx)) {
829 mss2_decode_end(avctx);
832 ff_mss2dsp_init(&ctx->dsp);
833 ff_qpeldsp_init(&ctx->qdsp);
835 avctx->pix_fmt = c->free_colours == 127 ? AV_PIX_FMT_RGB555
838 ctx->last_pic = av_frame_alloc();
839 if (!ctx->last_pic) {
840 mss2_decode_end(avctx);
841 return AVERROR(ENOMEM);
847 AVCodec ff_mss2_decoder = {
849 .long_name = NULL_IF_CONFIG_SMALL("MS Windows Media Video V9 Screen"),
850 .type = AVMEDIA_TYPE_VIDEO,
851 .id = AV_CODEC_ID_MSS2,
852 .priv_data_size = sizeof(MSS2Context),
853 .init = mss2_decode_init,
854 .close = mss2_decode_end,
855 .decode = mss2_decode_frame,
856 .capabilities = CODEC_CAP_DR1,