MSS2 decoder

Signed-off-by: Kostya Shishkov <kostya.shishkov@gmail.com>

16 files changed, 1461 insertions, 216 deletions

diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index adbe1a3e55..7b3dc22d7c 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -271,6 +271,7 @@ OBJS-$(CONFIG_MSMPEG4V3_ENCODER)       += msmpeg4.o msmpeg4enc.o msmpeg4data.o \
 OBJS-$(CONFIG_MSRLE_DECODER)           += msrle.o msrledec.o
 OBJS-$(CONFIG_MSA1_DECODER)            += mss3.o mss34dsp.o
 OBJS-$(CONFIG_MSS1_DECODER)            += mss1.o mss12.o
+OBJS-$(CONFIG_MSS2_DECODER)            += mss2.o mss12.o mss2dsp.o
 OBJS-$(CONFIG_MSVIDEO1_DECODER)        += msvideo1.o
 OBJS-$(CONFIG_MSZH_DECODER)            += lcldec.o
 OBJS-$(CONFIG_MTS2_DECODER)            += mss4.o mss34dsp.o
diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c
index f7187d1cf7..c6df818477 100644
--- a/libavcodec/allcodecs.c
+++ b/libavcodec/allcodecs.c
@@ -160,6 +160,7 @@ void avcodec_register_all(void)
     REGISTER_ENCDEC  (MSMPEG4V3, msmpeg4v3);
     REGISTER_DECODER (MSRLE, msrle);
     REGISTER_DECODER (MSS1, mss1);
+    REGISTER_DECODER (MSS2, mss2);
     REGISTER_DECODER (MSVIDEO1, msvideo1);
     REGISTER_DECODER (MSZH, mszh);
     REGISTER_DECODER (MTS2, mts2);
diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h
index 742a2a39cb..8a091bd524 100644
--- a/libavcodec/avcodec.h
+++ b/libavcodec/avcodec.h
@@ -264,6 +264,7 @@ enum AVCodecID {
     AV_CODEC_ID_TSCC2,
     AV_CODEC_ID_MTS2,
     AV_CODEC_ID_CLLC,
+    AV_CODEC_ID_MSS2,
 
     /* various PCM "codecs" */
     AV_CODEC_ID_FIRST_AUDIO = 0x10000,     ///< A dummy id pointing at the start of audio codecs
diff --git a/libavcodec/codec_desc.c b/libavcodec/codec_desc.c
index 513a15c090..51fc1714d1 100644
--- a/libavcodec/codec_desc.c
+++ b/libavcodec/codec_desc.c
@@ -1200,6 +1200,13 @@ static const AVCodecDescriptor codec_descriptors[] = {
         .long_name = NULL_IF_CONFIG_SMALL("Canopus Lossless Codec"),
         .props     = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSLESS,
     },
+    {
+        .id        = AV_CODEC_ID_MSS2,
+        .type      = AVMEDIA_TYPE_VIDEO,
+        .name      = "mss2",
+        .long_name = NULL_IF_CONFIG_SMALL("MS Windows Media Video V9 Screen"),
+        .props     = AV_CODEC_PROP_INTRA_ONLY | AV_CODEC_PROP_LOSSY,
+    },
 
     /* various PCM "codecs" */
     {
diff --git a/libavcodec/h263dec.c b/libavcodec/h263dec.c
index 6bb6204d97..8e6085beeb 100644
--- a/libavcodec/h263dec.c
+++ b/libavcodec/h263dec.c
@@ -58,7 +58,10 @@ av_cold int ff_h263_decode_init(AVCodecContext *avctx)
     s->quant_precision=5;
     s->decode_mb= ff_h263_decode_mb;
     s->low_delay= 1;
-    avctx->pix_fmt= avctx->get_format(avctx, avctx->codec->pix_fmts);
+    if (avctx->codec->id == AV_CODEC_ID_MSS2)
+        avctx->pix_fmt = PIX_FMT_YUV420P;
+    else
+        avctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts);
     s->unrestricted_mv= 1;
 
     /* select sub codec */
@@ -93,6 +96,7 @@ av_cold int ff_h263_decode_init(AVCodecContext *avctx)
     case AV_CODEC_ID_WMV3:
     case AV_CODEC_ID_VC1IMAGE:
     case AV_CODEC_ID_WMV3IMAGE:
+    case AV_CODEC_ID_MSS2:
         s->h263_pred = 1;
         s->msmpeg4_version=6;
         avctx->chroma_sample_location = AVCHROMA_LOC_LEFT;
diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c
index 13e2accfbd..e9aff3b316 100644
--- a/libavcodec/mpegvideo.c
+++ b/libavcodec/mpegvideo.c
@@ -226,10 +226,11 @@ void ff_copy_picture(Picture *dst, Picture *src)
  */
 static void free_frame_buffer(MpegEncContext *s, Picture *pic)
 {
-    /* Windows Media Image codecs allocate internal buffers with different
-     * dimensions; ignore user defined callbacks for these
-     */
-    if (s->codec_id != AV_CODEC_ID_WMV3IMAGE && s->codec_id != AV_CODEC_ID_VC1IMAGE)
+    /* WM Image / Screen codecs allocate internal buffers with different
+     * dimensions / colorspaces; ignore user-defined callbacks for these. */
+    if (s->codec_id != AV_CODEC_ID_WMV3IMAGE &&
+        s->codec_id != AV_CODEC_ID_VC1IMAGE  &&
+        s->codec_id != AV_CODEC_ID_MSS2)
         ff_thread_release_buffer(s->avctx, &pic->f);
     else
         avcodec_default_release_buffer(s->avctx, &pic->f);
@@ -254,7 +255,9 @@ static int alloc_frame_buffer(MpegEncContext *s, Picture *pic)
         }
     }
 
-    if (s->codec_id != AV_CODEC_ID_WMV3IMAGE && s->codec_id != AV_CODEC_ID_VC1IMAGE)
+    if (s->codec_id != AV_CODEC_ID_WMV3IMAGE &&
+        s->codec_id != AV_CODEC_ID_VC1IMAGE  &&
+        s->codec_id != AV_CODEC_ID_MSS2)
         r = ff_thread_get_buffer(s->avctx, &pic->f);
     else
         r = avcodec_default_get_buffer(s->avctx, &pic->f);
diff --git a/libavcodec/mss1.c b/libavcodec/mss1.c
index fe8ee10840..ada479eee7 100644
--- a/libavcodec/mss1.c
+++ b/libavcodec/mss1.c
@@ -24,14 +24,13 @@
  * Microsoft Screen 1 (aka Windows Media Video V7 Screen) decoder
  */
 
-#include "libavutil/intfloat.h"
-#include "libavutil/intreadwrite.h"
 #include "avcodec.h"
 #include "mss12.h"
 
 typedef struct MSS1Context {
     MSS12Context   ctx;
     AVFrame        pic;
+    SliceContext   sc[2];
 } MSS1Context;
 
 static void arith_normalise(ArithCoder *c)
@@ -56,24 +55,11 @@ static void arith_normalise(ArithCoder *c)
         c->low   <<= 1;
         c->high  <<= 1;
         c->high   |= 1;
-        c->value  |= get_bits1(c->gb);
+        c->value  |= get_bits1(c->gbc.gb);
     }
 }
 
-static int arith_get_bit(ArithCoder *c)
-{
-    int range = c->high - c->low + 1;
-    int bit   = (((c->value - c->low) << 1) + 1) / range;
-
-    if (bit)
-        c->low += range >> 1;
-    else
-        c->high = c->low + (range >> 1) - 1;
-
-    arith_normalise(c);
-
-    return bit;
-}
+ARITH_GET_BIT()
 
 static int arith_get_bits(ArithCoder *c, int bits)
 {
@@ -118,40 +104,27 @@ static int arith_get_prob(ArithCoder *c, int *probs)
     return sym;
 }
 
-static int arith_get_model_sym(ArithCoder *c, Model *m)
-{
-    int idx, val;
-
-    idx = arith_get_prob(c, m->cum_prob);
-
-    val = m->idx2sym[idx];
-    ff_mss12_model_update(m, idx);
-
-    arith_normalise(c);
-
-    return val;
-}
+ARITH_GET_MODEL_SYM()
 
 static void arith_init(ArithCoder *c, GetBitContext *gb)
 {
-    c->low   = 0;
-    c->high  = 0xFFFF;
-    c->value = get_bits(gb, 16);
-    c->gb    = gb;
-
+    c->low           = 0;
+    c->high          = 0xFFFF;
+    c->value         = get_bits(gb, 16);
+    c->gbc.gb        = gb;
     c->get_model_sym = arith_get_model_sym;
     c->get_number    = arith_get_number;
 }
 
-static int decode_pal(MSS1Context *ctx, ArithCoder *acoder)
+static int decode_pal(MSS12Context *ctx, ArithCoder *acoder)
 {
     int i, ncol, r, g, b;
-    uint32_t *pal = ctx->ctx.pal + 256 - ctx->ctx.free_colours;
+    uint32_t *pal = ctx->pal + 256 - ctx->free_colours;
 
-    if (!ctx->ctx.free_colours)
+    if (!ctx->free_colours)
         return 0;
 
-    ncol = arith_get_number(acoder, ctx->ctx.free_colours + 1);
+    ncol = arith_get_number(acoder, ctx->free_colours + 1);
     for (i = 0; i < ncol; i++) {
         r = arith_get_bits(acoder, 8);
         g = arith_get_bits(acoder, 8);
@@ -167,7 +140,8 @@ static int mss1_decode_frame(AVCodecContext *avctx, void *data, int *data_size,
 {
     const uint8_t *buf = avpkt->data;
     int buf_size = avpkt->size;
-    MSS1Context *c = avctx->priv_data;
+    MSS1Context *ctx = avctx->priv_data;
+    MSS12Context *c = &ctx->ctx;
     GetBitContext gb;
     ArithCoder acoder;
     int pal_changed = 0;
@@ -176,37 +150,37 @@ static int mss1_decode_frame(AVCodecContext *avctx, void *data, int *data_size,
     init_get_bits(&gb, buf, buf_size * 8);
     arith_init(&acoder, &gb);
 
-    c->pic.reference    = 3;
-    c->pic.buffer_hints = FF_BUFFER_HINTS_VALID | FF_BUFFER_HINTS_PRESERVE |
-                          FF_BUFFER_HINTS_REUSABLE;
-    if ((ret = avctx->reget_buffer(avctx, &c->pic)) < 0) {
+    ctx->pic.reference    = 3;
+    ctx->pic.buffer_hints = FF_BUFFER_HINTS_VALID | FF_BUFFER_HINTS_READABLE |
+                            FF_BUFFER_HINTS_PRESERVE | FF_BUFFER_HINTS_REUSABLE;
+    if ((ret = avctx->reget_buffer(avctx, &ctx->pic)) < 0) {
         av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n");
         return ret;
     }
 
-    c->ctx.pic_start  = c->pic.data[0] + c->pic.linesize[0] * (avctx->height - 1);
-    c->ctx.pic_stride = -c->pic.linesize[0];
-    c->ctx.keyframe   = !arith_get_bit(&acoder);
-    if (c->ctx.keyframe) {
-        ff_mss12_codec_reset(&c->ctx);
-        pal_changed      = decode_pal(c, &acoder);
-        c->pic.key_frame = 1;
-        c->pic.pict_type = AV_PICTURE_TYPE_I;
+    c->pal_pic    =  ctx->pic.data[0] + ctx->pic.linesize[0] * (avctx->height - 1);
+    c->pal_stride = -ctx->pic.linesize[0];
+    c->keyframe   = !arith_get_bit(&acoder);
+    if (c->keyframe) {
+        ff_mss12_codec_reset(c);
+        pal_changed        = decode_pal(c, &acoder);
+        ctx->pic.key_frame = 1;
+        ctx->pic.pict_type = AV_PICTURE_TYPE_I;
     } else {
-        if (c->ctx.corrupted)
+        if (c->corrupted)
             return AVERROR_INVALIDDATA;
-        c->pic.key_frame = 0;
-        c->pic.pict_type = AV_PICTURE_TYPE_P;
+        ctx->pic.key_frame = 0;
+        ctx->pic.pict_type = AV_PICTURE_TYPE_P;
     }
-    c->ctx.corrupted = ff_mss12_decode_rect(&c->ctx, &acoder, 0, 0,
-                                            avctx->width, avctx->height);
-    if (c->ctx.corrupted)
+    c->corrupted = ff_mss12_decode_rect(&c->sc[0], &acoder, 0, 0,
+                                        avctx->width, avctx->height);
+    if (c->corrupted)
         return AVERROR_INVALIDDATA;
-    memcpy(c->pic.data[1], c->ctx.pal, AVPALETTE_SIZE);
-    c->pic.palette_has_changed = pal_changed;
+    memcpy(ctx->pic.data[1], c->pal, AVPALETTE_SIZE);
+    ctx->pic.palette_has_changed = pal_changed;
 
     *data_size = sizeof(AVFrame);
-    *(AVFrame*)data = c->pic;
+    *(AVFrame*)data = ctx->pic;
 
     /* always report that the buffer was completely consumed */
     return buf_size;
@@ -219,16 +193,16 @@ static av_cold int mss1_decode_init(AVCodecContext *avctx)
     c->ctx.avctx       = avctx;
     avctx->coded_frame = &c->pic;
 
-    return ff_mss12_decode_init(avctx, 0);
+    return ff_mss12_decode_init(&c->ctx, 0);
 }
 
 static av_cold int mss1_decode_end(AVCodecContext *avctx)
 {
-    MSS1Context * const c = avctx->priv_data;
+    MSS1Context * const ctx = avctx->priv_data;
 
-    if (c->pic.data[0])
-        avctx->release_buffer(avctx, &c->pic);
-    ff_mss12_decode_end(avctx);
+    if (ctx->pic.data[0])
+        avctx->release_buffer(avctx, &ctx->pic);
+    ff_mss12_decode_end(&ctx->ctx);
 
     return 0;
 }
diff --git a/libavcodec/mss12.c b/libavcodec/mss12.c
index 8b4f52bf5f..f0f23c0f8e 100644
--- a/libavcodec/mss12.c
+++ b/libavcodec/mss12.c
@@ -47,12 +47,8 @@ static int model_calc_threshold(Model *m)
 {
     int thr;
 
-    if (m->thr_weight == -1) {
-        thr = 2 * m->weights[m->num_syms] - 1;
-        thr = ((thr >> 1) + 4 * m->cum_prob[0]) / thr;
-    } else {
-        thr = m->num_syms * m->thr_weight;
-    }
+    thr = 2 * m->weights[m->num_syms] - 1;
+    thr = ((thr >> 1) + 4 * m->cum_prob[0]) / thr;
 
     return FFMIN(thr, 0x3FFF);
 }
@@ -78,7 +74,7 @@ static av_cold void model_init(Model *m, int num_syms, int thr_weight)
 {
     m->num_syms   = num_syms;
     m->thr_weight = thr_weight;
-    m->threshold  = model_calc_threshold(m);
+    m->threshold  = num_syms * thr_weight;
     model_reset(m);
 }
 
@@ -87,7 +83,7 @@ static void model_rescale_weights(Model *m)
     int i;
     int cum_prob;
 
-    if (m->thr_weight == -1)
+    if (m->thr_weight == THRESH_ADAPTIVE)
         m->threshold = model_calc_threshold(m);
     while (m->cum_prob[0] > m->threshold) {
         cum_prob = 0;
@@ -129,8 +125,14 @@ static void pixctx_reset(PixContext *ctx)
 {
     int i, j, k;
 
-    for (i = 0; i < ctx->cache_size; i++)
-        ctx->cache[i] = i;
+    if (!ctx->special_initial_cache)
+        for (i = 0; i < ctx->cache_size; i++)
+            ctx->cache[i] = i;
+    else {
+        ctx->cache[0] = 1;
+        ctx->cache[1] = 2;
+        ctx->cache[2] = 4;
+    }
 
     model_reset(&ctx->cache_model);
     model_reset(&ctx->full_model);
@@ -141,27 +143,23 @@ static void pixctx_reset(PixContext *ctx)
                 model_reset(&ctx->sec_models[i][j][k]);
 }
 
-static av_cold void pixctx_init(PixContext *ctx, int cache_size)
+static av_cold void pixctx_init(PixContext *ctx, int cache_size,
+                                int full_model_syms, int special_initial_cache)
 {
     int i, j, k;
 
-    ctx->cache_size = cache_size + 4;
-    ctx->num_syms   = cache_size;
-
-    for (i = 0; i < ctx->cache_size; i++)
-        ctx->cache[i] = i;
+    ctx->cache_size            = cache_size + 4;
+    ctx->num_syms              = cache_size;
+    ctx->special_initial_cache = special_initial_cache;
 
     model_init(&ctx->cache_model, ctx->num_syms + 1, THRESH_LOW);
-    model_init(&ctx->full_model, 256, THRESH_HIGH);
+    model_init(&ctx->full_model, full_model_syms, THRESH_HIGH);
 
-    for (i = 0; i < 4; i++) {
-        for (j = 0; j < sec_order_sizes[i]; j++) {
-            for (k = 0; k < 4; k++) {
+    for (i = 0; i < 4; i++)
+        for (j = 0; j < sec_order_sizes[i]; j++)
+            for (k = 0; k < 4; k++)
                 model_init(&ctx->sec_models[i][j][k], 2 + i,
                            i ? THRESH_LOW : THRESH_ADAPTIVE);
-            }
-        }
-    }
 }
 
 static int decode_top_left_pixel(ArithCoder *acoder, PixContext *pctx)
@@ -196,7 +194,6 @@ static int decode_pixel(ArithCoder *acoder, PixContext *pctx,
     if (val < pctx->num_syms) {
         int idx, j;
 
-
         idx = 0;
         for (i = 0; i < pctx->cache_size; i++) {
             for (j = 0; j < num_ngb; j++)
@@ -309,195 +306,288 @@ static int decode_pixel_in_context(ArithCoder *acoder, PixContext *pctx,
         break;
     }
 
-    pix = acoder->get_model_sym(acoder, &pctx->sec_models[nlen - 1][layer][sub]);
+    pix = acoder->get_model_sym(acoder,
+                                &pctx->sec_models[nlen - 1][layer][sub]);
     if (pix < nlen)
         return ref_pix[pix];
     else
         return decode_pixel(acoder, pctx, ref_pix, nlen);
 }
 
-static int decode_region(MSS12Context *ctx, ArithCoder *acoder, uint8_t *dst,
+static int decode_region(ArithCoder *acoder, uint8_t *dst, uint8_t *rgb_pic,
                          int x, int y, int width, int height, int stride,
-                         PixContext *pctx)
+                         int rgb_stride, PixContext *pctx, const uint32_t *pal)
 {
-    int i, j;
+    int i, j, p;
+    uint8_t *rgb_dst = rgb_pic + x * 3 + y * rgb_stride;
 
     dst += x + y * stride;
 
-    dst[0] = decode_top_left_pixel(acoder, pctx);
     for (j = 0; j < height; j++) {
         for (i = 0; i < width; i++) {
             if (!i && !j)
-                continue;
+                p = decode_top_left_pixel(acoder, pctx);
+            else
+                p = decode_pixel_in_context(acoder, pctx, dst + i, stride,
+                                            i, j, width - i - 1);
+            dst[i] = p;
 
-            dst[i] = decode_pixel_in_context(acoder, pctx, dst + i, stride,
-                                             i, j, width - i - 1);
+            if (rgb_pic)
+                AV_WB24(rgb_dst + i * 3, pal[p]);
         }
-        dst += stride;
+        dst     += stride;
+        rgb_dst += rgb_stride;
     }
 
     return 0;
 }
 
-static int decode_region_masked(MSS12Context *ctx, ArithCoder *acoder,
+static void copy_rectangles(MSS12Context const *c,
+                            int x, int y, int width, int height)
+{
+    int j;
+
+    if (c->last_rgb_pic)
+        for (j = y; j < y + height; j++) {
+            memcpy(c->rgb_pic      + j * c->rgb_stride + x * 3,
+                   c->last_rgb_pic + j * c->rgb_stride + x * 3,
+                   width * 3);
+            memcpy(c->pal_pic      + j * c->pal_stride + x,
+                   c->last_pal_pic + j * c->pal_stride + x,
+                   width);
+        }
+}
+
+static int motion_compensation(MSS12Context const *c,
+                               int x, int y, int width, int height)
+{
+    if (x + c->mvX < 0 || x + c->mvX + width  > c->avctx->width  ||
+        y + c->mvY < 0 || y + c->mvY + height > c->avctx->height ||
+        !c->rgb_pic)
+        return -1;
+    else {
+        uint8_t *dst     = c->pal_pic + x     + y * c->pal_stride;
+        uint8_t *rgb_dst = c->rgb_pic + x * 3 + y * c->rgb_stride;
+        uint8_t *src;
+        uint8_t *rgb_src;
+        int j;
+        x += c->mvX;
+        y += c->mvY;
+        if (c->last_rgb_pic) {
+            src     = c->last_pal_pic + x +     y * c->pal_stride;
+            rgb_src = c->last_rgb_pic + x * 3 + y * c->rgb_stride;
+        } else {
+            src     = c->pal_pic + x     + y * c->pal_stride;
+            rgb_src = c->rgb_pic + x * 3 + y * c->rgb_stride;
+        }
+        for (j = 0; j < height; j++) {
+            memmove(dst, src, width);
+            memmove(rgb_dst, rgb_src, width * 3);
+            dst     += c->pal_stride;
+            src     += c->pal_stride;
+            rgb_dst += c->rgb_stride;
+            rgb_src += c->rgb_stride;
+        }
+    }
+    return 0;
+}
+
+static int decode_region_masked(MSS12Context const *c, ArithCoder *acoder,
                                 uint8_t *dst, int stride, uint8_t *mask,
                                 int mask_stride, int x, int y,
                                 int width, int height,
                                 PixContext *pctx)
 {
-    int i, j;
+    int i, j, p;
+    uint8_t *rgb_dst = c->rgb_pic + x * 3 + y * c->rgb_stride;
 
     dst  += x + y * stride;
     mask += x + y * mask_stride;
 
-    if (mask[0] == 0xFF)
-        dst[0] = decode_top_left_pixel(acoder, pctx);
     for (j = 0; j < height; j++) {
         for (i = 0; i < width; i++) {
-            if (!i && !j || mask[i] != 0xFF)
-                continue;
-
-            dst[i] = decode_pixel_in_context(acoder, pctx, dst + i, stride,
-                                             i, j, width - i - 1);
+            if (c->avctx->err_recognition & AV_EF_EXPLODE &&
+                ( c->rgb_pic && mask[i] != 0x01 && mask[i] != 0x02 && mask[i] != 0x04 ||
+                 !c->rgb_pic && mask[i] != 0x80 && mask[i] != 0xFF))
+                return -1;
+
+            if (mask[i] == 0x02) {
+                copy_rectangles(c, x + i, y + j, 1, 1);
+            } else if (mask[i] == 0x04) {
+                if (motion_compensation(c, x + i, y + j, 1, 1))
+                    return -1;
+            } else if (mask[i] != 0x80) {
+                if (!i && !j)
+                    p = decode_top_left_pixel(acoder, pctx);
+                else
+                    p = decode_pixel_in_context(acoder, pctx, dst + i, stride,
+                                                i, j, width - i - 1);
+                dst[i] = p;
+                if (c->rgb_pic)
+                    AV_WB24(rgb_dst + i * 3, c->pal[p]);
+            }
         }
-        dst  += stride;
-        mask += mask_stride;
+        dst     += stride;
+        mask    += mask_stride;
+        rgb_dst += c->rgb_stride;
     }
 
     return 0;
 }
 
-static av_cold void codec_init(MSS12Context *ctx)
+static av_cold void codec_init(MSS12Context *c, int version)
 {
-    model_init(&ctx->intra_region, 2, THRESH_ADAPTIVE);
-    model_init(&ctx->inter_region, 2, THRESH_ADAPTIVE);
-    model_init(&ctx->split_mode,   3, THRESH_HIGH);
-    model_init(&ctx->edge_mode,    2, THRESH_HIGH);
-    model_init(&ctx->pivot,        3, THRESH_LOW);
-    pixctx_init(&ctx->intra_pix_ctx, 8);
-    pixctx_init(&ctx->inter_pix_ctx, 2);
-    ctx->corrupted = 1;
+    int i;
+    for (i = 0; i < (c->slice_split ? 2 : 1); i++) {
+        c->sc[i].c = c;
+        model_init(&c->sc[i].intra_region, 2, THRESH_ADAPTIVE);
+        model_init(&c->sc[i].inter_region, 2, THRESH_ADAPTIVE);
+        model_init(&c->sc[i].split_mode,   3, THRESH_HIGH);
+        model_init(&c->sc[i].edge_mode,    2, THRESH_HIGH);
+        model_init(&c->sc[i].pivot,        3, THRESH_LOW);
+
+        pixctx_init(&c->sc[i].intra_pix_ctx, 8, c->full_model_syms, 0);
+
+        pixctx_init(&c->sc[i].inter_pix_ctx, version ? 3 : 2,
+                    c->full_model_syms, version ? 1 : 0);
+    }
+    c->corrupted = 1;
 }
 
-void ff_mss12_codec_reset(MSS12Context *ctx)
+void ff_mss12_codec_reset(MSS12Context *c)
 {
-    model_reset(&ctx->intra_region);
-    model_reset(&ctx->inter_region);
-    model_reset(&ctx->split_mode);
-    model_reset(&ctx->edge_mode);
-    model_reset(&ctx->pivot);
-    pixctx_reset(&ctx->intra_pix_ctx);
-    pixctx_reset(&ctx->inter_pix_ctx);
-
-    ctx->corrupted = 0;
+    int i;
+    for (i = 0; i < (c->slice_split ? 2 : 1); i++) {
+        model_reset(&c->sc[i].intra_region);
+        model_reset(&c->sc[i].inter_region);
+        model_reset(&c->sc[i].split_mode);
+        model_reset(&c->sc[i].edge_mode);
+        model_reset(&c->sc[i].pivot);
+        pixctx_reset(&c->sc[i].intra_pix_ctx);
+        pixctx_reset(&c->sc[i].inter_pix_ctx);
+    }
+
+    c->corrupted = 0;
 }
 
-static int decode_pivot(MSS12Context *ctx, ArithCoder *acoder, int base)
+static int decode_pivot(SliceContext *sc, ArithCoder *acoder, int base)
 {
     int val, inv;
 
-    inv = acoder->get_model_sym(acoder, &ctx->edge_mode);
-    val = acoder->get_model_sym(acoder, &ctx->pivot) + 1;
+    inv = acoder->get_model_sym(acoder, &sc->edge_mode);
+    val = acoder->get_model_sym(acoder, &sc->pivot) + 1;
 
     if (val > 2) {
-        if ((base + 1) / 2 - 2 <= 0) {
-            ctx->corrupted = 1;
-            return 0;
-        }
+        if ((base + 1) / 2 - 2 <= 0)
+            return -1;
+
         val = acoder->get_number(acoder, (base + 1) / 2 - 2) + 3;
     }
 
-    if (val == base) {
-        ctx->corrupted = 1;
-        return 0;
-    }
+    if (val >= base)
+        return -1;
 
     return inv ? base - val : val;
 }
 
-static int decode_region_intra(MSS12Context *ctx, ArithCoder *acoder,
+static int decode_region_intra(SliceContext *sc, ArithCoder *acoder,
                                int x, int y, int width, int height)
 {
+    MSS12Context const *c = sc->c;
     int mode;
 
-    mode = acoder->get_model_sym(acoder, &ctx->intra_region);
+    mode = acoder->get_model_sym(acoder, &sc->intra_region);
 
     if (!mode) {
-        int i, pix;
-        int stride = ctx->pic_stride;
-        uint8_t *dst = ctx->pic_start + x + y * stride;
-
-        pix = decode_top_left_pixel(acoder, &ctx->intra_pix_ctx);
-        for (i = 0; i < height; i++, dst += stride)
+        int i, j, pix, rgb_pix;
+        int stride       = c->pal_stride;
+        int rgb_stride   = c->rgb_stride;
+        uint8_t *dst     = c->pal_pic + x     + y * stride;
+        uint8_t *rgb_dst = c->rgb_pic + x * 3 + y * rgb_stride;
+
+        pix     = decode_top_left_pixel(acoder, &sc->intra_pix_ctx);
+        rgb_pix = c->pal[pix];
+        for (i = 0; i < height; i++, dst += stride, rgb_dst += rgb_stride) {
             memset(dst, pix, width);
+            if (c->rgb_pic)
+                for (j = 0; j < width * 3; j += 3)
+                    AV_WB24(rgb_dst + j, rgb_pix);
+        }
     } else {
-        return decode_region(ctx, acoder, ctx->pic_start,
-                             x, y, width, height, ctx->pic_stride,
-                             &ctx->intra_pix_ctx);
+        return decode_region(acoder, c->pal_pic, c->rgb_pic,
+                             x, y, width, height, c->pal_stride, c->rgb_stride,
+                             &sc->intra_pix_ctx, &c->pal[0]);
     }
 
     return 0;
 }
 
-static int decode_region_inter(MSS12Context *ctx, ArithCoder *acoder,
+static int decode_region_inter(SliceContext *sc, ArithCoder *acoder,
                                int x, int y, int width, int height)
 {
+    MSS12Context const *c = sc->c;
     int mode;
 
-    mode = acoder->get_model_sym(acoder, &ctx->inter_region);
+    mode = acoder->get_model_sym(acoder, &sc->inter_region);
 
     if (!mode) {
-        mode = decode_top_left_pixel(acoder, &ctx->inter_pix_ctx);
-        if (mode != 0xFF) {
-            return 0;
-        } else {
-            return decode_region_intra(ctx, acoder, x, y, width, height);
-        }
+        mode = decode_top_left_pixel(acoder, &sc->inter_pix_ctx);
+
+        if (c->avctx->err_recognition & AV_EF_EXPLODE &&
+            ( c->rgb_pic && mode != 0x01 && mode != 0x02 && mode != 0x04 ||
+             !c->rgb_pic && mode != 0x80 && mode != 0xFF))
+            return -1;
+
+        if (mode == 0x02)
+            copy_rectangles(c, x, y, width, height);
+        else if (mode == 0x04)
+            return motion_compensation(c, x, y, width, height);
+        else if (mode != 0x80)
+            return decode_region_intra(sc, acoder, x, y, width, height);
     } else {
-        if (decode_region(ctx, acoder, ctx->mask,
-                          x, y, width, height, ctx->mask_linesize,
-                          &ctx->inter_pix_ctx) < 0)
+        if (decode_region(acoder, c->mask, NULL,
+                          x, y, width, height, c->mask_stride, 0,
+                          &sc->inter_pix_ctx, &c->pal[0]) < 0)
             return -1;
-        return decode_region_masked(ctx, acoder, ctx->pic_start,
-                                    ctx->pic_stride, ctx->mask,
-                                    ctx->mask_linesize,
+        return decode_region_masked(c, acoder, c->pal_pic,
+                                    c->pal_stride, c->mask,
+                                    c->mask_stride,
                                     x, y, width, height,
-                                    &ctx->intra_pix_ctx);
+                                    &sc->intra_pix_ctx);
     }
 
     return 0;
 }
 
-int ff_mss12_decode_rect(MSS12Context *ctx, ArithCoder *acoder,
+int ff_mss12_decode_rect(SliceContext *sc, ArithCoder *acoder,
                          int x, int y, int width, int height)
 {
     int mode, pivot;
 
-    if (ctx->corrupted)
-        return -1;
-
-    mode = acoder->get_model_sym(acoder, &ctx->split_mode);
+    mode = acoder->get_model_sym(acoder, &sc->split_mode);
 
     switch (mode) {
     case SPLIT_VERT:
-        pivot = decode_pivot(ctx, acoder, height);
-        if (ff_mss12_decode_rect(ctx, acoder, x, y, width, pivot))
+        if ((pivot = decode_pivot(sc, acoder, height)) < 1)
             return -1;
-        if (ff_mss12_decode_rect(ctx, acoder, x, y + pivot, width, height - pivot))
+        if (ff_mss12_decode_rect(sc, acoder, x, y, width, pivot))
+            return -1;
+        if (ff_mss12_decode_rect(sc, acoder, x, y + pivot, width, height - pivot))
             return -1;
         break;
     case SPLIT_HOR:
-        pivot = decode_pivot(ctx, acoder, width);
-        if (ff_mss12_decode_rect(ctx, acoder, x, y, pivot, height))
+        if ((pivot = decode_pivot(sc, acoder, width)) < 1)
+            return -1;
+        if (ff_mss12_decode_rect(sc, acoder, x, y, pivot, height))
             return -1;
-        if (ff_mss12_decode_rect(ctx, acoder, x + pivot, y, width - pivot, height))
+        if (ff_mss12_decode_rect(sc, acoder, x + pivot, y, width - pivot, height))
             return -1;
         break;
     case SPLIT_NONE:
-        if (ctx->keyframe)
-            return decode_region_intra(ctx, acoder, x, y, width, height);
+        if (sc->c->keyframe)
+            return decode_region_intra(sc, acoder, x, y, width, height);
         else
-            return decode_region_inter(ctx, acoder, x, y, width, height);
+            return decode_region_inter(sc, acoder, x, y, width, height);
     default:
         return -1;
     }
@@ -505,13 +595,11 @@ int ff_mss12_decode_rect(MSS12Context *ctx, ArithCoder *acoder,
     return 0;
 }
 
-av_cold int ff_mss12_decode_init(AVCodecContext *avctx, int version)
+av_cold int ff_mss12_decode_init(MSS12Context *c, int version)
 {
-    MSS12Context * const c = avctx->priv_data;
+    AVCodecContext *avctx = c->avctx;
     int i;
 
-    c->avctx = avctx;
-
     if (avctx->extradata_size < 52 + 256 * 3) {
         av_log(avctx, AV_LOG_ERROR, "Insufficient extradata size %d\n",
                avctx->extradata_size);
@@ -526,9 +614,23 @@ av_cold int ff_mss12_decode_init(AVCodecContext *avctx, int version)
         return AVERROR_INVALIDDATA;
     }
 
+    avctx->coded_width  = AV_RB32(avctx->extradata + 20);
+    avctx->coded_height = AV_RB32(avctx->extradata + 24);
+    if (avctx->coded_width > 4096 || avctx->coded_height > 4096) {
+        av_log(avctx, AV_LOG_ERROR, "Frame dimensions %dx%d too large",
+               avctx->coded_width, avctx->coded_height);
+        return AVERROR_INVALIDDATA;
+    }
+
     av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d\n",
            AV_RB32(avctx->extradata + 4), AV_RB32(avctx->extradata + 8));
-    c->free_colours     = AV_RB32(avctx->extradata + 48);
+    if (version != AV_RB32(avctx->extradata + 4) > 1) {
+        av_log(avctx, AV_LOG_ERROR,
+               "Header version doesn't match codec tag\n");
+        return -1;
+    }
+
+    c->free_colours = AV_RB32(avctx->extradata + 48);
     if ((unsigned)c->free_colours > 256) {
         av_log(avctx, AV_LOG_ERROR,
                "Incorrect number of changeable palette entries: %d\n",
@@ -536,8 +638,6 @@ av_cold int ff_mss12_decode_init(AVCodecContext *avctx, int version)
         return AVERROR_INVALIDDATA;
     }
     av_log(avctx, AV_LOG_DEBUG, "%d free colour(s)\n", c->free_colours);
-    avctx->coded_width  = AV_RB32(avctx->extradata + 20);
-    avctx->coded_height = AV_RB32(avctx->extradata + 24);
 
     av_log(avctx, AV_LOG_DEBUG, "Display dimensions %dx%d\n",
            AV_RB32(avctx->extradata + 12), AV_RB32(avctx->extradata + 16));
@@ -554,27 +654,53 @@ av_cold int ff_mss12_decode_init(AVCodecContext *avctx, int version)
     av_log(avctx, AV_LOG_DEBUG, "Max. seek time %g ms\n",
            av_int2float(AV_RB32(avctx->extradata + 44)));
 
-    for (i = 0; i < 256; i++)
-        c->pal[i] = AV_RB24(avctx->extradata + 52 + i * 3);
+    if (version) {
+        if (avctx->extradata_size < 60 + 256 * 3) {
+            av_log(avctx, AV_LOG_ERROR,
+                   "Insufficient extradata size %d for v2\n",
+                   avctx->extradata_size);
+            return AVERROR_INVALIDDATA;
+        }
+
+        c->slice_split = AV_RB32(avctx->extradata + 52);
+        av_log(avctx, AV_LOG_DEBUG, "Slice split %d\n", c->slice_split);
 
-    avctx->pix_fmt = PIX_FMT_PAL8;
+        c->full_model_syms = AV_RB32(avctx->extradata + 56);
+        if (c->full_model_syms < 2 || c->full_model_syms > 256) {
+            av_log(avctx, AV_LOG_ERROR,
+                   "Incorrect number of used colours %d\n",
+                   c->full_model_syms);
+            return AVERROR_INVALIDDATA;
+        }
+        av_log(avctx, AV_LOG_DEBUG, "Used colours %d\n",
+               c->full_model_syms);
+    } else {
+        c->slice_split     = 0;
+        c->full_model_syms = 256;
+    }
 
-    c->mask_linesize = FFALIGN(avctx->width, 16);
-    c->mask          = av_malloc(c->mask_linesize * avctx->height);
+    for (i = 0; i < 256; i++)
+        c->pal[i] = AV_RB24(avctx->extradata + 52 +
+                            (version ? 8 : 0) + i * 3);
+
+    c->mask_stride = FFALIGN(avctx->width, 16);
+    c->mask        = av_malloc(c->mask_stride * avctx->height);
     if (!c->mask) {
         av_log(avctx, AV_LOG_ERROR, "Cannot allocate mask plane\n");
         return AVERROR(ENOMEM);
     }
 
-    codec_init(c);
+    avctx->pix_fmt = version ? c->free_colours == 127 ? PIX_FMT_RGB555
+                                                      : PIX_FMT_RGB24
+                             : PIX_FMT_PAL8;
+
+    codec_init(c, version);
 
     return 0;
 }
 
-av_cold int ff_mss12_decode_end(AVCodecContext *avctx)
+av_cold int ff_mss12_decode_end(MSS12Context *c)
 {
-    MSS12Context * const c = avctx->priv_data;
-
     av_freep(&c->mask);
 
     return 0;
diff --git a/libavcodec/mss12.h b/libavcodec/mss12.h
index 383d86c904..93d1f6146f 100644
--- a/libavcodec/mss12.h
+++ b/libavcodec/mss12.h
@@ -26,8 +26,10 @@
 #ifndef AVCODEC_MSS12_H
 #define AVCODEC_MSS12_H
 
+#include "libavutil/intreadwrite.h"
 #include "avcodec.h"
 #include "get_bits.h"
+#include "bytestream.h"
 
 #define MODEL_MIN_SYMS    2
 #define MODEL_MAX_SYMS  256
@@ -46,7 +48,10 @@ typedef struct Model {
 
 typedef struct ArithCoder {
     int low, high, value;
-    GetBitContext *gb;
+    union {
+        GetBitContext *gb;
+        GetByteContext *gB;
+    } gbc;
     int (*get_model_sym)(struct ArithCoder *c, Model *m);
     int (*get_number)   (struct ArithCoder *c, int n);
 } ArithCoder;
@@ -56,28 +61,77 @@ typedef struct PixContext {
     uint8_t cache[12];
     Model cache_model, full_model;
     Model sec_models[4][8][4];
+    int special_initial_cache;
 } PixContext;
 
+struct MSS12Context;
+
+typedef struct SliceContext {
+    struct MSS12Context *c;
+    Model      intra_region, inter_region;
+    Model      pivot, edge_mode, split_mode;
+    PixContext intra_pix_ctx, inter_pix_ctx;
+} SliceContext;
+
 typedef struct MSS12Context {
     AVCodecContext *avctx;
-    uint8_t        *pic_start;
-    int            pic_stride;
-    uint8_t        *mask;
-    int            mask_linesize;
     uint32_t       pal[256];
+    uint8_t        *pal_pic;
+    uint8_t        *last_pal_pic;
+    int            pal_stride;
+    uint8_t        *mask;
+    int            mask_stride;
+    uint8_t        *rgb_pic;
+    uint8_t        *last_rgb_pic;
+    int            rgb_stride;
     int            free_colours;
     int            keyframe;
     Model          intra_region, inter_region;
     Model          pivot, edge_mode, split_mode;
     PixContext     intra_pix_ctx, inter_pix_ctx;
+    int            mvX, mvY;
     int            corrupted;
+    int            slice_split;
+    int            full_model_syms;
+    SliceContext   sc[2];
 } MSS12Context;
 
-int ff_mss12_decode_rect(MSS12Context *ctx, ArithCoder *acoder,
+int ff_mss12_decode_rect(SliceContext *ctx, ArithCoder *acoder,
                          int x, int y, int width, int height);
 void ff_mss12_model_update(Model *m, int val);
 void ff_mss12_codec_reset(MSS12Context *ctx);
-av_cold int ff_mss12_decode_init(AVCodecContext *avctx, int version);
-av_cold int ff_mss12_decode_end(AVCodecContext *avctx);
+av_cold int ff_mss12_decode_init(MSS12Context *ctx, int version);
+av_cold int ff_mss12_decode_end(MSS12Context *ctx);
+
+#define ARITH_GET_BIT(VERSION)                                          \
+static int arith ## VERSION ## _get_bit(ArithCoder *c)                  \
+{                                                                       \
+    int range = c->high - c->low + 1;                                   \
+    int bit   = (((c->value - c->low) << 1) + 1) / range;               \
+                                                                        \
+    if (bit)                                                            \
+        c->low += range >> 1;                                           \
+    else                                                                \
+        c->high = c->low + (range >> 1) - 1;                            \
+                                                                        \
+    arith ## VERSION ## _normalise(c);                                  \
+                                                                        \
+    return bit;                                                         \
+}
+
+#define ARITH_GET_MODEL_SYM(VERSION)                                    \
+static int arith ## VERSION ## _get_model_sym(ArithCoder *c, Model *m)  \
+{                                                                       \
+    int idx, val;                                                       \
+                                                                        \
+    idx = arith ## VERSION ## _get_prob(c, m->cum_prob);                \
+                                                                        \
+    val = m->idx2sym[idx];                                              \
+    ff_mss12_model_update(m, idx);                                      \
+                                                                        \
+    arith ## VERSION ## _normalise(c);                                  \
+                                                                        \
+    return val;                                                         \
+}
 
 #endif /* AVCODEC_MSS12_H */
diff --git a/libavcodec/mss2.c b/libavcodec/mss2.c
new file mode 100644
index 0000000000..c0c47dc8d3
--- /dev/null
+++ b/libavcodec/mss2.c
@@ -0,0 +1,860 @@
+/*
+ * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder
+ */
+
+#include "libavutil/avassert.h"
+#include "msmpeg4data.h"
+#include "vc1.h"
+#include "mss12.h"
+#include "mss2dsp.h"
+
+typedef struct MSS2Context {
+    VC1Context     v;
+    int            split_position;
+    AVFrame        pic;
+    AVFrame        last_pic;
+    MSS12Context   c;
+    MSS2DSPContext dsp;
+    SliceContext   sc[2];
+} MSS2Context;
+
+static void arith2_normalise(ArithCoder *c)
+{
+    while ((c->high >> 15) - (c->low >> 15) < 2) {
+        if ((c->low ^ c->high) & 0x10000) {
+            c->high  ^= 0x8000;
+            c->value ^= 0x8000;
+            c->low   ^= 0x8000;
+        }
+        c->high  = c->high  << 8 & 0xFFFFFF | 0xFF;
+        c->value = c->value << 8 & 0xFFFFFF | bytestream2_get_byte(c->gbc.gB);
+        c->low   = c->low   << 8 & 0xFFFFFF;
+    }
+}
+
+ARITH_GET_BIT(2)
+
+/* L. Stuiver and A. Moffat: "Piecewise Integer Mapping for Arithmetic Coding."
+ * In Proc. 8th Data Compression Conference (DCC '98), pp. 3-12, Mar. 1998 */
+
+static int arith2_get_scaled_value(int value, int n, int range)
+{
+    int split = (n << 1) - range;
+
+    if (value > split)
+        return split + (value - split >> 1);
+    else
+        return value;
+}
+
+static void arith2_rescale_interval(ArithCoder *c, int range,
+                                    int low, int high, int n)
+{
+    int split = (n << 1) - range;
+
+    if (high > split)
+        c->high = split + (high - split << 1);
+    else
+        c->high = high;
+
+    c->high += c->low - 1;
+
+    if (low > split)
+        c->low += split + (low - split << 1);
+    else
+        c->low += low;
+}
+
+static int arith2_get_number(ArithCoder *c, int n)
+{
+    int range = c->high - c->low + 1;
+    int scale = av_log2(range) - av_log2(n);
+    int val;
+
+    if (n << scale > range)
+        scale--;
+
+    n <<= scale;
+
+    val = arith2_get_scaled_value(c->value - c->low, n, range) >> scale;
+
+    arith2_rescale_interval(c, range, val << scale, (val + 1) << scale, n);
+
+    arith2_normalise(c);
+
+    return val;
+}
+
+static int arith2_get_prob(ArithCoder *c, int *probs)
+{
+    int range = c->high - c->low + 1, n = *probs;
+    int scale = av_log2(range) - av_log2(n);
+    int i     = 0, val;
+
+    if (n << scale > range)
+        scale--;
+
+    n <<= scale;
+
+    val = arith2_get_scaled_value(c->value - c->low, n, range) >> scale;
+    while (probs[++i] > val) ;
+
+    arith2_rescale_interval(c, range,
+                            probs[i] << scale, probs[i - 1] << scale, n);
+
+    return i;
+}
+
+ARITH_GET_MODEL_SYM(2)
+
+static int arith2_get_consumed_bytes(ArithCoder *c)
+{
+    int diff = (c->high >> 16) - (c->low >> 16);
+    int bp   = bytestream2_tell(c->gbc.gB) - 3 << 3;
+    int bits = 1;
+
+    while (!(diff & 0x80)) {
+        bits++;
+        diff <<= 1;
+    }
+
+    return (bits + bp + 7 >> 3) + ((c->low >> 16) + 1 == c->high >> 16);
+}
+
+static void arith2_init(ArithCoder *c, GetByteContext *gB)
+{
+    c->low           = 0;
+    c->high          = 0xFFFFFF;
+    c->value         = bytestream2_get_be24(gB);
+    c->gbc.gB        = gB;
+    c->get_model_sym = arith2_get_model_sym;
+    c->get_number    = arith2_get_number;
+}
+
+static int decode_pal_v2(MSS12Context *ctx, const uint8_t *buf, int buf_size)
+{
+    int i, ncol;
+    uint32_t *pal = ctx->pal + 256 - ctx->free_colours;
+
+    if (!ctx->free_colours)
+        return 0;
+
+    ncol = *buf++;
+    if (buf_size < 2 + ncol * 3)
+        return -1;
+    for (i = 0; i < ncol; i++)
+        *pal++ = AV_RB24(buf + 3 * i);
+
+    return 1 + ncol * 3;
+}
+
+static int decode_555(GetByteContext *gB, uint16_t *dst, int stride,
+                      int keyframe, int w, int h)
+{
+    int last_symbol = 0, repeat = 0, prev_avail = 0;
+
+    if (!keyframe) {
+        int x, y, endx, endy, t;
+
+#define READ_PAIR(a, b)                 \
+    a  = bytestream2_get_byte(gB) << 4; \
+    t  = bytestream2_get_byte(gB);      \
+    a |= t >> 4;                        \
+    b  = (t & 0xF) << 8;                \
+    b |= bytestream2_get_byte(gB);      \
+
+        READ_PAIR(x, endx)
+        READ_PAIR(y, endy)
+
+        if (endx >= w || endy >= h || x > endx || y > endy)
+            return -1;
+        dst += x + stride * y;
+        w    = endx - x + 1;
+        h    = endy - y + 1;
+        if (y)
+            prev_avail = 1;
+    }
+
+    do {
+        uint16_t *p = dst;
+        do {
+            if (repeat-- < 1) {
+                int b = bytestream2_get_byte(gB);
+                if (b < 128)
+                    last_symbol = b << 8 | bytestream2_get_byte(gB);
+                else if (b > 129) {
+                    repeat = 0;
+                    while (b-- > 130)
+                        repeat = (repeat << 8) + bytestream2_get_byte(gB) + 1;
+                    if (last_symbol == -2) {
+                        int skip = FFMIN((unsigned)repeat, dst + w - p);
+                        repeat -= skip;
+                        p      += skip;
+                    }
+                } else
+                    last_symbol = 127 - b;
+            }
+            if (last_symbol >= 0)
+                *p = last_symbol;
+            else if (last_symbol == -1 && prev_avail)
+                *p = *(p - stride);
+        } while (++p < dst + w);
+        dst       += stride;
+        prev_avail = 1;
+    } while (--h);
+
+    return 0;
+}
+
+static int decode_rle(GetBitContext *gb, uint8_t *pal_dst, int pal_stride,
+                      uint8_t *rgb_dst, int rgb_stride, uint32_t *pal,
+                      int keyframe, int kf_slipt, int slice, int w, int h)
+{
+    uint8_t bits[270] = { 0 };
+    uint32_t codes[270];
+    VLC vlc;
+
+    int current_length = 0, read_codes = 0, next_code = 0, current_codes = 0;
+    int remaining_codes, surplus_codes, i;
+
+    const int alphabet_size = 270 - keyframe;
+
+    int last_symbol = 0, repeat = 0, prev_avail = 0;
+
+    if (!keyframe) {
+        int x, y, clipw, cliph;
+
+        x     = get_bits(gb, 12);
+        y     = get_bits(gb, 12);
+        clipw = get_bits(gb, 12) + 1;
+        cliph = get_bits(gb, 12) + 1;
+
+        if (x + clipw > w || y + cliph > h)
+            return AVERROR_INVALIDDATA;
+        pal_dst += pal_stride * y + x;
+        rgb_dst += rgb_stride * y + x * 3;
+        w        = clipw;
+        h        = cliph;
+        if (y)
+            prev_avail = 1;
+    } else {
+        if (slice > 0) {
+            pal_dst   += pal_stride * kf_slipt;
+            rgb_dst   += rgb_stride * kf_slipt;
+            prev_avail = 1;
+            h         -= kf_slipt;
+        } else
+            h = kf_slipt;
+    }
+
+    /* read explicit codes */
+    do {
+        while (current_codes--) {
+            int symbol = get_bits(gb, 8);
+            if (symbol >= 204 - keyframe)
+                symbol += 14 - keyframe;
+            else if (symbol > 189)
+                symbol = get_bits1(gb) + (symbol << 1) - 190;
+            if (bits[symbol])
+                return AVERROR_INVALIDDATA;
+            bits[symbol]  = current_length;
+            codes[symbol] = next_code++;
+            read_codes++;
+        }
+        current_length++;
+        next_code     <<= 1;
+        remaining_codes = (1 << current_length) - next_code;
+        current_codes   = get_bits(gb, av_ceil_log2(remaining_codes + 1));
+        if (current_length > 22 || current_codes > remaining_codes)
+            return AVERROR_INVALIDDATA;
+    } while (current_codes != remaining_codes);
+
+    remaining_codes = alphabet_size - read_codes;
+
+    /* determine the minimum length to fit the rest of the alphabet */
+    while ((surplus_codes = (2 << current_length) -
+                            (next_code << 1) - remaining_codes) < 0) {
+        current_length++;
+        next_code <<= 1;
+    }
+
+    /* add the rest of the symbols lexicographically */
+    for (i = 0; i < alphabet_size; i++)
+        if (!bits[i]) {
+            if (surplus_codes-- == 0) {
+                current_length++;
+                next_code <<= 1;
+            }
+            bits[i]  = current_length;
+            codes[i] = next_code++;
+        }
+
+    if (next_code != 1 << current_length)
+        return AVERROR_INVALIDDATA;
+
+    if (i = init_vlc(&vlc, 9, alphabet_size, bits, 1, 1, codes, 4, 4, 0))
+        return i;
+
+    /* frame decode */
+    do {
+        uint8_t *pp = pal_dst;
+        uint8_t *rp = rgb_dst;
+        do {
+            if (repeat-- < 1) {
+                int b = get_vlc2(gb, vlc.table, 9, 3);
+                if (b < 256)
+                    last_symbol = b;
+                else if (b < 268) {
+                    b -= 256;
+                    if (b == 11)
+                        b = get_bits(gb, 4) + 10;
+
+                    if (!b)
+                        repeat = 0;
+                    else
+                        repeat = get_bits(gb, b);
+
+                    while (b--)
+                        repeat += 1 << b;
+
+                    if (last_symbol == -2) {
+                        int skip = FFMIN(repeat, pal_dst + w - pp);
+                        repeat -= skip;
+                        pp     += skip;
+                        rp     += skip * 3;
+                    }
+                } else
+                    last_symbol = 267 - b;
+            }
+            if (last_symbol >= 0) {
+                *pp = last_symbol;
+                AV_WB24(rp, pal[last_symbol]);
+            } else if (last_symbol == -1 && prev_avail) {
+                *pp = *(pp - pal_stride);
+                memcpy(rp, rp - rgb_stride, 3);
+            }
+            rp += 3;
+        } while (++pp < pal_dst + w);
+        pal_dst   += pal_stride;
+        rgb_dst   += rgb_stride;
+        prev_avail = 1;
+    } while (--h);
+
+    ff_free_vlc(&vlc);
+    return 0;
+}
+
+static int decode_wmv9(AVCodecContext *avctx, const uint8_t *buf, int buf_size,
+                       int x, int y, int w, int h, int wmv9_mask)
+{
+    MSS2Context *ctx  = avctx->priv_data;
+    MSS12Context *c   = &ctx->c;
+    VC1Context *v     = avctx->priv_data;
+    MpegEncContext *s = &v->s;
+    AVFrame *f;
+
+    ff_mpeg_flush(avctx);
+
+    if (s->current_picture_ptr == NULL || s->current_picture_ptr->f.data[0]) {
+        int i = ff_find_unused_picture(s, 0);
+        if (i < 0)
+            return -1;
+        s->current_picture_ptr = &s->picture[i];
+    }
+
+    init_get_bits(&s->gb, buf, buf_size * 8);
+
+    s->loop_filter = avctx->skip_loop_filter < AVDISCARD_ALL;
+
+    if (ff_vc1_parse_frame_header(v, &s->gb) == -1) {
+        av_log(v->s.avctx, AV_LOG_ERROR, "header error\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    if (s->pict_type != AV_PICTURE_TYPE_I) {
+        av_log(v->s.avctx, AV_LOG_ERROR, "expected I-frame\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    avctx->pix_fmt = PIX_FMT_YUV420P;
+
+    if (ff_MPV_frame_start(s, avctx) < 0) {
+        av_log(v->s.avctx, AV_LOG_ERROR, "ff_MPV_frame_start error\n");
+        avctx->pix_fmt = PIX_FMT_RGB24;
+        return -1;
+    }
+
+    ff_er_frame_start(s);
+
+    v->bits = buf_size * 8;
+
+    v->end_mb_x = (w + 15) >> 4;
+    s->end_mb_y = (h + 15) >> 4;
+    if (v->respic & 1)
+        v->end_mb_x = v->end_mb_x + 1 >> 1;
+    if (v->respic & 2)
+        s->end_mb_y = s->end_mb_y + 1 >> 1;
+
+    ff_vc1_decode_blocks(v);
+
+    ff_er_frame_end(s);
+
+    ff_MPV_frame_end(s);
+
+    f = &s->current_picture.f;
+
+    if (v->respic == 3) {
+        ctx->dsp.upsample_plane(f->data[0], f->linesize[0], w,      h);
+        ctx->dsp.upsample_plane(f->data[1], f->linesize[1], w >> 1, h >> 1);
+        ctx->dsp.upsample_plane(f->data[2], f->linesize[2], w >> 1, h >> 1);
+    } else if (v->respic)
+        av_log_ask_for_sample(v->s.avctx,
+                              "Asymmetric WMV9 rectangle subsampling\n");
+
+    av_assert0(f->linesize[1] == f->linesize[2]);
+
+    if (wmv9_mask != -1)
+        ctx->dsp.mss2_blit_wmv9_masked(c->rgb_pic + y * c->rgb_stride + x * 3,
+                                       c->rgb_stride, wmv9_mask,
+                                       c->pal_pic + y * c->pal_stride + x,
+                                       c->pal_stride,
+                                       f->data[0], f->linesize[0],
+                                       f->data[1], f->data[2], f->linesize[1],
+                                       w, h);
+    else
+        ctx->dsp.mss2_blit_wmv9(c->rgb_pic + y * c->rgb_stride + x * 3,
+                                c->rgb_stride,
+                                f->data[0], f->linesize[0],
+                                f->data[1], f->data[2], f->linesize[1],
+                                w, h);
+
+    avctx->pix_fmt = PIX_FMT_RGB24;
+
+    return 0;
+}
+
+typedef struct Rectangle {
+    int coded, x, y, w, h;
+} Rectangle;
+
+#define MAX_WMV9_RECTANGLES 20
+#define ARITH2_PADDING 2
+
+static int mss2_decode_frame(AVCodecContext *avctx, void *data, int *data_size,
+                             AVPacket *avpkt)
+{
+    const uint8_t *buf = avpkt->data;
+    int buf_size       = avpkt->size;
+    MSS2Context *ctx = avctx->priv_data;
+    MSS12Context *c  = &ctx->c;
+    GetBitContext gb;
+    GetByteContext gB;
+    ArithCoder acoder;
+
+    int keyframe, has_wmv9, has_mv, is_rle, is_555, ret;
+
+    Rectangle wmv9rects[MAX_WMV9_RECTANGLES], *r;
+    int used_rects = 0, i, implicit_rect, av_uninit(wmv9_mask);
+
+    av_assert0(FF_INPUT_BUFFER_PADDING_SIZE >=
+               ARITH2_PADDING + (MIN_CACHE_BITS + 7) / 8);
+
+    init_get_bits(&gb, buf, buf_size * 8);
+
+    if (keyframe = get_bits1(&gb))
+        skip_bits(&gb, 7);
+    has_wmv9 = get_bits1(&gb);
+    has_mv   = keyframe ? 0 : get_bits1(&gb);
+    is_rle   = get_bits1(&gb);
+    is_555   = is_rle && get_bits1(&gb);
+    if (c->slice_split > 0)
+        ctx->split_position = c->slice_split;
+    else if (c->slice_split < 0) {
+        if (get_bits1(&gb)) {
+            if (get_bits1(&gb)) {
+                if (get_bits1(&gb))
+                    ctx->split_position = get_bits(&gb, 16);
+                else
+                    ctx->split_position = get_bits(&gb, 12);
+            } else
+                ctx->split_position = get_bits(&gb, 8) << 4;
+        } else {
+            if (keyframe)
+                ctx->split_position = avctx->height / 2;
+        }
+    } else
+        ctx->split_position = avctx->height;
+
+    if (c->slice_split && (ctx->split_position < 1 - is_555 ||
+                           ctx->split_position > avctx->height - 1))
+        return AVERROR_INVALIDDATA;
+
+    align_get_bits(&gb);
+    buf      += get_bits_count(&gb) >> 3;
+    buf_size -= get_bits_count(&gb) >> 3;
+
+    if (buf_size < 1)
+        return AVERROR_INVALIDDATA;
+
+    if (is_555 && (has_wmv9 || has_mv || c->slice_split && ctx->split_position))
+        return AVERROR_INVALIDDATA;
+
+    avctx->pix_fmt = is_555 ? PIX_FMT_RGB555 : PIX_FMT_RGB24;
+    if (ctx->pic.data[0] && ctx->pic.format != avctx->pix_fmt)
+        avctx->release_buffer(avctx, &ctx->pic);
+
+    if (has_wmv9) {
+        bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
+        arith2_init(&acoder, &gB);
+
+        implicit_rect = !arith2_get_bit(&acoder);
+
+        while (arith2_get_bit(&acoder)) {
+            if (used_rects == MAX_WMV9_RECTANGLES)
+                return AVERROR_INVALIDDATA;
+            r = &wmv9rects[used_rects];
+            if (!used_rects)
+                r->x = arith2_get_number(&acoder, avctx->width);
+            else
+                r->x = arith2_get_number(&acoder, avctx->width -
+                                         wmv9rects[used_rects - 1].x) +
+                       wmv9rects[used_rects - 1].x;
+            r->y = arith2_get_number(&acoder, avctx->height);
+            r->w = arith2_get_number(&acoder, avctx->width  - r->x) + 1;
+            r->h = arith2_get_number(&acoder, avctx->height - r->y) + 1;
+            used_rects++;
+        }
+
+        if (implicit_rect && used_rects) {
+            av_log(avctx, AV_LOG_ERROR, "implicit_rect && used_rects > 0\n");
+            return AVERROR_INVALIDDATA;
+        }
+
+        if (implicit_rect) {
+            wmv9rects[0].x = 0;
+            wmv9rects[0].y = 0;
+            wmv9rects[0].w = avctx->width;
+            wmv9rects[0].h = avctx->height;
+
+            used_rects = 1;
+        }
+        for (i = 0; i < used_rects; i++) {
+            if (!implicit_rect && arith2_get_bit(&acoder)) {
+                av_log(avctx, AV_LOG_ERROR, "Unexpected grandchildren\n");
+                return AVERROR_INVALIDDATA;
+            }
+            if (!i) {
+                wmv9_mask = arith2_get_bit(&acoder) - 1;
+                if (!wmv9_mask)
+                    wmv9_mask = arith2_get_number(&acoder, 256);
+            }
+            wmv9rects[i].coded = arith2_get_number(&acoder, 2);
+        }
+
+        buf      += arith2_get_consumed_bytes(&acoder);
+        buf_size -= arith2_get_consumed_bytes(&acoder);
+        if (buf_size < 1)
+            return AVERROR_INVALIDDATA;
+    }
+
+    c->mvX = c->mvY = 0;
+    if (keyframe && !is_555) {
+        if ((i = decode_pal_v2(c, buf, buf_size)) < 0)
+            return AVERROR_INVALIDDATA;
+        buf      += i;
+        buf_size -= i;
+    } else if (has_mv) {
+        buf      += 4;
+        buf_size -= 4;
+        if (buf_size < 1)
+            return AVERROR_INVALIDDATA;
+        c->mvX = AV_RB16(buf - 4) - avctx->width;
+        c->mvY = AV_RB16(buf - 2) - avctx->height;
+    }
+
+    if (c->mvX < 0 || c->mvY < 0) {
+        FFSWAP(AVFrame, ctx->pic, ctx->last_pic);
+        FFSWAP(uint8_t *, c->pal_pic, c->last_pal_pic);
+
+        if (ctx->pic.data[0])
+            avctx->release_buffer(avctx, &ctx->pic);
+
+        ctx->pic.reference    = 3;
+        ctx->pic.buffer_hints = FF_BUFFER_HINTS_VALID    |
+                                FF_BUFFER_HINTS_READABLE |
+                                FF_BUFFER_HINTS_PRESERVE |
+                                FF_BUFFER_HINTS_REUSABLE;
+
+        if ((ret = avctx->get_buffer(avctx, &ctx->pic)) < 0) {
+            av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
+            return ret;
+        }
+
+        if (ctx->last_pic.data[0]) {
+            av_assert0(ctx->pic.linesize[0] == ctx->last_pic.linesize[0]);
+            c->last_rgb_pic = ctx->last_pic.data[0] +
+                              ctx->last_pic.linesize[0] * (avctx->height - 1);
+        } else {
+            av_log(avctx, AV_LOG_ERROR, "Missing keyframe\n");
+            return -1;
+        }
+    } else {
+        if (ctx->last_pic.data[0])
+            avctx->release_buffer(avctx, &ctx->last_pic);
+
+        ctx->pic.reference    = 3;
+        ctx->pic.buffer_hints = FF_BUFFER_HINTS_VALID    |
+                                FF_BUFFER_HINTS_READABLE |
+                                FF_BUFFER_HINTS_PRESERVE |
+                                FF_BUFFER_HINTS_REUSABLE;
+
+        if ((ret = avctx->reget_buffer(avctx, &ctx->pic)) < 0) {
+            av_log(avctx, AV_LOG_ERROR, "reget_buffer() failed\n");
+            return ret;
+        }
+
+        c->last_rgb_pic = NULL;
+    }
+    c->rgb_pic    = ctx->pic.data[0] +
+                    ctx->pic.linesize[0] * (avctx->height - 1);
+    c->rgb_stride = -ctx->pic.linesize[0];
+
+    ctx->pic.key_frame = keyframe;
+    ctx->pic.pict_type = keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
+
+    if (is_555) {
+        bytestream2_init(&gB, buf, buf_size);
+
+        if (decode_555(&gB, (uint16_t *)c->rgb_pic, c->rgb_stride >> 1,
+                       keyframe, avctx->width, avctx->height))
+            return AVERROR_INVALIDDATA;
+
+        buf_size -= bytestream2_tell(&gB);
+    } else if (is_rle) {
+        init_get_bits(&gb, buf, buf_size * 8);
+        if (ret = decode_rle(&gb, c->pal_pic, c->pal_stride,
+                             c->rgb_pic, c->rgb_stride, c->pal, keyframe,
+                             ctx->split_position, 0,
+                             avctx->width, avctx->height))
+            return ret;
+        align_get_bits(&gb);
+
+        if (c->slice_split)
+            if (ret = decode_rle(&gb, c->pal_pic, c->pal_stride,
+                                 c->rgb_pic, c->rgb_stride, c->pal, keyframe,
+                                 ctx->split_position, 1,
+                                 avctx->width, avctx->height))
+                return ret;
+
+        align_get_bits(&gb);
+        buf      += get_bits_count(&gb) >> 3;
+        buf_size -= get_bits_count(&gb) >> 3;
+    } else {
+        if (keyframe)
+            ff_mss12_codec_reset(c);
+        else if (c->corrupted)
+            return AVERROR_INVALIDDATA;
+        bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
+        arith2_init(&acoder, &gB);
+        c->keyframe = keyframe;
+        if (c->corrupted = ff_mss12_decode_rect(&c->sc[0], &acoder, 0, 0,
+                                                avctx->width,
+                                                ctx->split_position))
+            return AVERROR_INVALIDDATA;
+
+        buf      += arith2_get_consumed_bytes(&acoder);
+        buf_size -= arith2_get_consumed_bytes(&acoder);
+        if (c->slice_split) {
+            if (buf_size < 1)
+                return AVERROR_INVALIDDATA;
+            bytestream2_init(&gB, buf, buf_size + ARITH2_PADDING);
+            arith2_init(&acoder, &gB);
+            if (c->corrupted = ff_mss12_decode_rect(&c->sc[1], &acoder, 0,
+                                                    ctx->split_position,
+                                                    avctx->width,
+                                                    avctx->height - ctx->split_position))
+                return AVERROR_INVALIDDATA;
+
+            buf      += arith2_get_consumed_bytes(&acoder);
+            buf_size -= arith2_get_consumed_bytes(&acoder);
+        }
+    }
+
+    if (has_wmv9) {
+        for (i = 0; i < used_rects; i++) {
+            int x = wmv9rects[i].x;
+            int y = wmv9rects[i].y;
+            int w = wmv9rects[i].w;
+            int h = wmv9rects[i].h;
+            if (wmv9rects[i].coded) {
+                int WMV9codedFrameSize;
+                if (buf_size < 4 || !(WMV9codedFrameSize = AV_RL24(buf)))
+                    return AVERROR_INVALIDDATA;
+                if (ret = decode_wmv9(avctx, buf + 3, buf_size - 3,
+                                      x, y, w, h, wmv9_mask))
+                    return ret;
+                buf      += WMV9codedFrameSize + 3;
+                buf_size -= WMV9codedFrameSize + 3;
+            } else {
+                uint8_t *dst = c->rgb_pic + y * c->rgb_stride + x * 3;
+                if (wmv9_mask != -1) {
+                    ctx->dsp.mss2_gray_fill_masked(dst, c->rgb_stride,
+                                                   wmv9_mask,
+                                                   c->pal_pic + y * c->pal_stride + x,
+                                                   c->pal_stride,
+                                                   w, h);
+                } else {
+                    do {
+                        memset(dst, 0x80, w * 3);
+                        dst += c->rgb_stride;
+                    } while (--h);
+                }
+            }
+        }
+    }
+
+    if (buf_size)
+        av_log(avctx, AV_LOG_WARNING, "buffer not fully consumed\n");
+
+    *data_size       = sizeof(AVFrame);
+    *(AVFrame *)data = ctx->pic;
+
+    return avpkt->size;
+}
+
+static av_cold int wmv9_init(AVCodecContext *avctx)
+{
+    VC1Context *v = avctx->priv_data;
+
+    v->s.avctx    = avctx;
+    avctx->flags |= CODEC_FLAG_EMU_EDGE;
+    v->s.flags   |= CODEC_FLAG_EMU_EDGE;
+
+    if (avctx->idct_algo == FF_IDCT_AUTO)
+        avctx->idct_algo = FF_IDCT_WMV2;
+
+    if (ff_vc1_init_common(v) < 0)
+        return -1;
+    ff_vc1dsp_init(&v->vc1dsp);
+
+    v->profile = PROFILE_MAIN;
+
+    v->zz_8x4     = ff_wmv2_scantableA;
+    v->zz_4x8     = ff_wmv2_scantableB;
+    v->res_y411   = 0;
+    v->res_sprite = 0;
+
+    v->frmrtq_postproc = 7;
+    v->bitrtq_postproc = 31;
+
+    v->res_x8          = 0;
+    v->multires        = 0;
+    v->res_fasttx      = 1;
+
+    v->fastuvmc        = 0;
+
+    v->extended_mv     = 0;
+
+    v->dquant          = 1;
+    v->vstransform     = 1;
+
+    v->res_transtab    = 0;
+
+    v->overlap         = 0;
+
+    v->s.resync_marker = 0;
+    v->rangered        = 0;
+
+    v->s.max_b_frames = avctx->max_b_frames = 0;
+    v->quantizer_mode = 0;
+
+    v->finterpflag = 0;
+
+    v->res_rtm_flag = 1;
+
+    ff_vc1_init_transposed_scantables(v);
+
+    if (ff_msmpeg4_decode_init(avctx) < 0 ||
+        ff_vc1_decode_init_alloc_tables(v) < 0)
+        return -1;
+
+    /* error concealment */
+    v->s.me.qpel_put = v->s.dsp.put_qpel_pixels_tab;
+    v->s.me.qpel_avg = v->s.dsp.avg_qpel_pixels_tab;
+
+    return 0;
+}
+
+static av_cold int mss2_decode_end(AVCodecContext *avctx)
+{
+    MSS2Context *const ctx = avctx->priv_data;
+
+    if (ctx->pic.data[0])
+        avctx->release_buffer(avctx, &ctx->pic);
+    if (ctx->last_pic.data[0])
+        avctx->release_buffer(avctx, &ctx->last_pic);
+
+    ff_mss12_decode_end(&ctx->c);
+    av_freep(&ctx->c.pal_pic);
+    av_freep(&ctx->c.last_pal_pic);
+    ff_vc1_decode_end(avctx);
+
+    return 0;
+}
+
+static av_cold int mss2_decode_init(AVCodecContext *avctx)
+{
+    MSS2Context * const ctx = avctx->priv_data;
+    MSS12Context *c = &ctx->c;
+    int ret;
+    c->avctx = avctx;
+    avctx->coded_frame = &ctx->pic;
+    if (ret = ff_mss12_decode_init(c, 1))
+        return ret;
+    c->pal_stride   = c->mask_stride;
+    c->pal_pic      = av_malloc(c->pal_stride * avctx->height);
+    c->last_pal_pic = av_malloc(c->pal_stride * avctx->height);
+    if (!c->pal_pic || !c->last_pal_pic) {
+        mss2_decode_end(avctx);
+        return AVERROR(ENOMEM);
+    }
+    if (ret = wmv9_init(avctx)) {
+        mss2_decode_end(avctx);
+        return ret;
+    }
+    ff_mss2dsp_init(&ctx->dsp);
+    return 0;
+}
+
+AVCodec ff_mss2_decoder = {
+    .name           = "mss2",
+    .type           = AVMEDIA_TYPE_VIDEO,
+    .id             = AV_CODEC_ID_MSS2,
+    .priv_data_size = sizeof(MSS2Context),
+    .init           = mss2_decode_init,
+    .close          = mss2_decode_end,
+    .decode         = mss2_decode_frame,
+    .capabilities   = CODEC_CAP_DR1,
+    .long_name      = NULL_IF_CONFIG_SMALL("MS Windows Media Video V9 Screen"),
+};
diff --git a/libavcodec/mss2dsp.c b/libavcodec/mss2dsp.c
new file mode 100644
index 0000000000..aa13577d48
--- /dev/null
+++ b/libavcodec/mss2dsp.c
@@ -0,0 +1,153 @@
+/*
+ * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder DSP routines
+ */
+
+#include "mss2dsp.h"
+#include "libavutil/common.h"
+
+static av_always_inline void mss2_blit_wmv9_template(uint8_t *dst,
+                                                     int dst_stride,
+                                                     int gray,
+                                                     int use_mask,
+                                                     int maskcolor,
+                                                     const uint8_t *mask,
+                                                     int mask_stride,
+                                                     const uint8_t *srcy,
+                                                     int srcy_stride,
+                                                     const uint8_t *srcu,
+                                                     const uint8_t *srcv,
+                                                     int srcuv_stride,
+                                                     int w, int h)
+{
+    int i, j, k, r = -1;
+    while (++r < h) {
+        for (i = 0, j = 0, k = 0; i < w; j += (i & 1), i++, k += 3) {
+            if (!use_mask || mask[i] == maskcolor) {
+                if (gray) {
+                    dst[k] = dst[k + 1] = dst[k + 2] = 0x80;
+                } else {
+                    int y = srcy[i];
+                    int u = srcu[j] - 128;
+                    int v = srcv[j] - 128;
+                    dst[k]     = av_clip_uint8(y + (             91881 * v + 32768 >> 16));
+                    dst[k + 1] = av_clip_uint8(y + (-22554 * u - 46802 * v + 32768 >> 16));
+                    dst[k + 2] = av_clip_uint8(y + (116130 * u             + 32768 >> 16));
+                }
+            }
+        }
+        mask +=  mask_stride;
+        dst  +=   dst_stride;
+        srcy +=  srcy_stride;
+        srcu += srcuv_stride * (r & 1);
+        srcv += srcuv_stride * (r & 1);
+    }
+}
+
+static void mss2_blit_wmv9_c(uint8_t *dst, int dst_stride,
+                             const uint8_t *srcy, int srcy_stride,
+                             const uint8_t *srcu, const uint8_t *srcv,
+                             int srcuv_stride, int w, int h)
+{
+    mss2_blit_wmv9_template(dst, dst_stride, 0, 0,
+                            0, NULL, 0,
+                            srcy, srcy_stride,
+                            srcu, srcv, srcuv_stride,
+                            w, h);
+}
+
+static void mss2_blit_wmv9_masked_c(uint8_t *dst, int dst_stride,
+                                    int maskcolor, const uint8_t *mask,
+                                    int mask_stride,
+                                    const uint8_t *srcy, int srcy_stride,
+                                    const uint8_t *srcu, const uint8_t *srcv,
+                                    int srcuv_stride, int w, int h)
+{
+    mss2_blit_wmv9_template(dst, dst_stride, 0, 1,
+                            maskcolor, mask, mask_stride,
+                            srcy, srcy_stride,
+                            srcu, srcv, srcuv_stride,
+                            w, h);
+}
+
+static void mss2_gray_fill_masked_c(uint8_t *dst, int dst_stride,
+                                    int maskcolor, const uint8_t *mask,
+                                    int mask_stride, int w, int h)
+{
+    mss2_blit_wmv9_template(dst, dst_stride, 1, 1,
+                            maskcolor, mask, mask_stride,
+                            NULL, 0,
+                            NULL, NULL, 0,
+                            w, h);
+}
+
+static void upsample_plane_c(uint8_t *plane, int plane_stride, int w, int h)
+{
+    uint8_t *src1, *src2, *dst1, *dst2, *p, a, b;
+    int i, j;
+
+    w += (w & 1);
+    h += (h & 1);
+
+    j = h - 1;
+
+    memcpy(plane + plane_stride *  j,
+           plane + plane_stride * (j >> 1),
+           w);
+
+    while ((j -= 2) > 0) {
+        dst1 = plane + plane_stride *  (j + 1);
+        dst2 = plane + plane_stride *   j;
+        src1 = plane + plane_stride * ((j + 1) >> 1);
+        src2 = plane + plane_stride * ( j      >> 1);
+
+        for (i = (w - 1) >> 1; i >= 0; i--) {
+            a = src1[i];
+            b = src2[i];
+            dst1[i] = (3 * a + b + 2) >> 2;
+            dst2[i] = (a + 3 * b + 2) >> 2;
+        }
+    }
+
+    for (j = h - 1; j >= 0; j--) {
+        p = plane + plane_stride * j;
+        i = w - 1;
+
+        p[i] = p[i >> 1];
+
+        while ((i -= 2) > 0) {
+            a = p[ i      >> 1];
+            b = p[(i + 1) >> 1];
+            p[i]     = (3 * a + b + 1) >> 2;
+            p[i + 1] = (a + 3 * b + 1) >> 2;
+        }
+    }
+}
+
+av_cold void ff_mss2dsp_init(MSS2DSPContext* dsp)
+{
+    dsp->mss2_blit_wmv9        = mss2_blit_wmv9_c;
+    dsp->mss2_blit_wmv9_masked = mss2_blit_wmv9_masked_c;
+    dsp->mss2_gray_fill_masked = mss2_gray_fill_masked_c;
+    dsp->upsample_plane        = upsample_plane_c;
+}
diff --git a/libavcodec/mss2dsp.h b/libavcodec/mss2dsp.h
new file mode 100644
index 0000000000..b3d67a1e57
--- /dev/null
+++ b/libavcodec/mss2dsp.h
@@ -0,0 +1,50 @@
+/*
+ * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * Microsoft Screen 2 (aka Windows Media Video V9 Screen) decoder DSP routines
+ */
+
+#ifndef AVCODEC_MSS2DSP_H
+#define AVCODEC_MSS2DSP_H
+
+#include "dsputil.h"
+
+typedef struct MSS2DSPContext {
+    void (*mss2_blit_wmv9)(uint8_t *dst, int dst_stride,
+                           const uint8_t *srcy, int srcy_stride,
+                           const uint8_t *srcu, const uint8_t *srcv,
+                           int srcuv_stride, int w, int h);
+    void (*mss2_blit_wmv9_masked)(uint8_t *dst, int dst_stride,
+                                  int maskcolor, const uint8_t *mask,
+                                  int mask_stride,
+                                  const uint8_t *srcy, int srcy_stride,
+                                  const uint8_t *srcu, const uint8_t *srcv,
+                                  int srcuv_stride, int w, int h);
+    void (*mss2_gray_fill_masked)(uint8_t *dst, int dst_stride,
+                                  int maskcolor, const uint8_t *mask,
+                                  int mask_stride, int w, int h);
+    void (*upsample_plane)(uint8_t *plane, int plane_stride, int w, int h);
+} MSS2DSPContext;
+
+av_cold void ff_mss2dsp_init(MSS2DSPContext* dsp);
+
+#endif /* AVCODEC_MSS2DSP_H */
diff --git a/libavcodec/vc1.c b/libavcodec/vc1.c
index 0e218af759..de75aac5e1 100644
--- a/libavcodec/vc1.c
+++ b/libavcodec/vc1.c
@@ -578,7 +578,12 @@ int ff_vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)
 
     if (v->finterpflag)
         v->interpfrm = get_bits1(gb);
-    skip_bits(gb, 2); //framecnt unused
+    if (v->s.avctx->codec->id == AV_CODEC_ID_MSS2)
+        v->respic   =
+        v->rangered =
+        v->multires = get_bits(gb, 2) == 1;
+    else
+        skip_bits(gb, 2); //framecnt unused
     v->rangeredfrm = 0;
     if (v->rangered)
         v->rangeredfrm = get_bits1(gb);
diff --git a/libavcodec/vc1.h b/libavcodec/vc1.h
index fe21f2f6b7..13011ae038 100644
--- a/libavcodec/vc1.h
+++ b/libavcodec/vc1.h
@@ -394,6 +394,8 @@ typedef struct VC1Context{
     uint8_t broken_link;         ///< Broken link flag (BROKEN_LINK syntax element)
     uint8_t closed_entry;        ///< Closed entry point flag (CLOSED_ENTRY syntax element)
 
+    int end_mb_x;                ///< Horizontal macroblock limit (used only by mss2)
+
     int parse_only;              ///< Context is used within parser
 
     int warn_interlaced;
diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c
index 63c0949ccd..6b2662e8f4 100644
--- a/libavcodec/vc1dec.c
+++ b/libavcodec/vc1dec.c
@@ -4348,10 +4348,10 @@ static void vc1_decode_i_blocks(VC1Context *v)
     s->mb_x = s->mb_y = 0;
     s->mb_intra         = 1;
     s->first_slice_line = 1;
-    for (s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
+    for (s->mb_y = 0; s->mb_y < s->end_mb_y; s->mb_y++) {
         s->mb_x = 0;
         ff_init_block_index(s);
-        for (; s->mb_x < s->mb_width; s->mb_x++) {
+        for (; s->mb_x < v->end_mb_x; s->mb_x++) {
             uint8_t *dst[6];
             ff_update_block_index(s);
             dst[0] = s->dest[0];
@@ -4438,7 +4438,10 @@ static void vc1_decode_i_blocks(VC1Context *v)
         s->first_slice_line = 0;
     }
     if (v->s.loop_filter)
-        ff_draw_horiz_band(s, (s->mb_height - 1) * 16, 16);
+        ff_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);
+
+    /* This is intentionally mb_height and not end_mb_y - unlike in advanced
+     * profile, these only differ are when decoding MSS2 rectangles. */
     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, ER_MB_END);
 }
 
@@ -5549,6 +5552,7 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data,
         ff_er_frame_start(s);
 
         v->bits = buf_size * 8;
+        v->end_mb_x = s->mb_width;
         if (v->field_mode) {
             uint8_t *tmp[2];
             s->current_picture.f.linesize[0] <<= 1;
diff --git a/libavcodec/version.h b/libavcodec/version.h
index 78f4ccf17f..b08f00d274 100644
--- a/libavcodec/version.h
+++ b/libavcodec/version.h
@@ -27,8 +27,8 @@
  */
 
 #define LIBAVCODEC_VERSION_MAJOR 54
-#define LIBAVCODEC_VERSION_MINOR 26
-#define LIBAVCODEC_VERSION_MICRO  1
+#define LIBAVCODEC_VERSION_MINOR 27
+#define LIBAVCODEC_VERSION_MICRO  0
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
                                                LIBAVCODEC_VERSION_MINOR, \