gst/ffmpegcolorspace/: Add conversions from/to NV12 and NV21 and conversions between those two formats. Fixes bug #53...

Original commit message from CVS:
Based on a patch by:
Björn Benderius <bjoern dot benderius at axis dot com>
* gst/ffmpegcolorspace/avcodec.h:
* gst/ffmpegcolorspace/gstffmpegcodecmap.c:
(gst_ffmpeg_pixfmt_to_caps), (gst_ffmpeg_caps_to_pixfmt),
(gst_ffmpegcsp_avpicture_fill):
* gst/ffmpegcolorspace/imgconvert.c: (nv12_to_nv21):
* gst/ffmpegcolorspace/imgconvert_template.h:
Add conversions from/to NV12 and NV21 and conversions between those
two formats. Fixes bug #532166.

ChangeLog

@@ -1,3 +1,17 @@
+2008-05-09  Sebastian Dröge  <slomo@circular-chaos.org>
+
+	Based on a patch by:
+	  Björn Benderius <bjoern dot benderius at axis dot com>
+
+	* gst/ffmpegcolorspace/avcodec.h:
+	* gst/ffmpegcolorspace/gstffmpegcodecmap.c:
+	(gst_ffmpeg_pixfmt_to_caps), (gst_ffmpeg_caps_to_pixfmt),
+	(gst_ffmpegcsp_avpicture_fill):
+	* gst/ffmpegcolorspace/imgconvert.c: (nv12_to_nv21):
+	* gst/ffmpegcolorspace/imgconvert_template.h:
+	Add conversions from/to NV12 and NV21 and conversions between those
+	two formats. Fixes bug #532166.
+
 2008-05-08  Edward Hervey  <edward.hervey@collabora.co.uk>
 
 	* gst/typefind/gsttypefindfunctions.c: (h264_video_type_find):

common

@@ -1 +1 @@
-Subproject commit ba3dd2882b1611f8115f9664e3b85e1fd956b53d
+Subproject commit dbf8f3aeceb6e57de097951a670cd853b4886ad8

gst/ffmpegcolorspace/avcodec.h

@@ -53,6 +53,8 @@ enum CodecType {
  */
 enum PixelFormat {
     PIX_FMT_YUV420P,   ///< Planar YUV 4:2:0 (1 Cr & Cb sample per 2x2 Y samples) (I420)
+    PIX_FMT_NV12,      ///< Packed YUV 4:2:0 (separate Y plane, interleaved Cb & Cr planes)
+    PIX_FMT_NV21,      ///< Packed YUV 4:2:0 (separate Y plane, interleaved Cb & Cr planes)
     PIX_FMT_YVU420P,   ///< Planar YUV 4:2:0 (1 Cb & Cr sample per 2x2 Y samples) (YV12)
     PIX_FMT_YUV422,    ///< Packed pixel, Y0 Cb Y1 Cr 
     PIX_FMT_RGB24,     ///< Packed pixel, 3 bytes per pixel, RGBRGB...
@@ -60,9 +62,9 @@ enum PixelFormat {
     PIX_FMT_YUV422P,   ///< Planar YUV 4:2:2 (1 Cr & Cb sample per 2x1 Y samples)
     PIX_FMT_YUV444P,   ///< Planar YUV 4:4:4 (1 Cr & Cb sample per 1x1 Y samples)
     PIX_FMT_RGBA32,    ///< Packed pixel, 4 bytes per pixel, BGRABGRA..., stored in cpu endianness
-    PIX_FMT_BGRA32,     ///< Packed pixel, 4 bytes per pixel, ARGBARGB...
+    PIX_FMT_BGRA32,    ///< Packed pixel, 4 bytes per pixel, ARGBARGB...
     PIX_FMT_ARGB32,    ///< Packed pixel, 4 bytes per pixel, ABGRABGR..., stored in cpu endianness
-    PIX_FMT_ABGR32,     ///< Packed pixel, 4 bytes per pixel, RGBARGBA...
+    PIX_FMT_ABGR32,    ///< Packed pixel, 4 bytes per pixel, RGBARGBA...
     PIX_FMT_RGB32,     ///< Packed pixel, 4 bytes per pixel, BGRxBGRx..., stored in cpu endianness
     PIX_FMT_xRGB32,    ///< Packed pixel, 4 bytes per pixel, xBGRxBGR..., stored in cpu endianness
     PIX_FMT_BGR32,     ///< Packed pixel, 4 bytes per pixel, xRGBxRGB...

gst/ffmpegcolorspace/gstffmpegcodecmap.c

@@ -167,6 +167,12 @@ gst_ffmpeg_pixfmt_to_caps (enum PixelFormat pix_fmt, AVCodecContext * context)
     case PIX_FMT_YUV420P:
       fmt = GST_MAKE_FOURCC ('I', '4', '2', '0');
       break;
+    case PIX_FMT_NV12:
+      fmt = GST_MAKE_FOURCC ('N', 'V', '1', '2');
+      break;
+    case PIX_FMT_NV21:
+      fmt = GST_MAKE_FOURCC ('N', 'V', '2', '1');
+      break;
     case PIX_FMT_YVU420P:
       fmt = GST_MAKE_FOURCC ('Y', 'V', '1', '2');
       break;
@@ -593,6 +599,12 @@ gst_ffmpeg_caps_to_pixfmt (const GstCaps * caps,
         case GST_MAKE_FOURCC ('I', '4', '2', '0'):
           context->pix_fmt = PIX_FMT_YUV420P;
           break;
+        case GST_MAKE_FOURCC ('N', 'V', '1', '2'):
+          context->pix_fmt = PIX_FMT_NV12;
+          break;
+        case GST_MAKE_FOURCC ('N', 'V', '2', '1'):
+          context->pix_fmt = PIX_FMT_NV21;
+          break;
         case GST_MAKE_FOURCC ('Y', 'V', '1', '2'):
           context->pix_fmt = PIX_FMT_YVU420P;
           break;
@@ -799,6 +811,22 @@ gst_ffmpegcsp_avpicture_fill (AVPicture * picture,
       picture->linesize[1] = stride2;
       picture->linesize[2] = stride2;
       return size + 2 * size2;
+    case PIX_FMT_NV12:
+    case PIX_FMT_NV21:
+      stride = GST_ROUND_UP_4 (width);
+      h2 = ROUND_UP_X (height, pinfo->y_chroma_shift);
+      size = stride * h2;
+      w2 = 2 * DIV_ROUND_UP_X (width, pinfo->x_chroma_shift);
+      stride2 = GST_ROUND_UP_4 (w2);
+      h2 = DIV_ROUND_UP_X (height, pinfo->y_chroma_shift);
+      size2 = stride2 * h2;
+      picture->data[0] = ptr;
+      picture->data[1] = picture->data[0] + size;
+      picture->data[2] = NULL;
+      picture->linesize[0] = stride;
+      picture->linesize[1] = stride2;
+      picture->linesize[2] = 0;
+      return size + size2;
     case PIX_FMT_RGB24:
     case PIX_FMT_BGR24:
       stride = GST_ROUND_UP_4 (width * 3);

gst/ffmpegcolorspace/imgconvert.c

@@ -64,6 +64,28 @@ static PixFmtInfo pix_fmt_info[PIX_FMT_NB] = {
         /* .y_chroma_shift = */ 1,
         /* .depth          = */ 8,
       },
+  /* [PIX_FMT_NV12] = */ {
+        /* .format         = */ PIX_FMT_NV12,
+        /* .name           = */ "nv12",
+        /* .nb_channels    = */ 2,
+        /* .color_type     = */ FF_COLOR_YUV,
+        /* .pixel_type     = */ FF_PIXEL_PACKED,
+        /* .is_alpha       = */ 0,
+        /* .x_chroma_shift = */ 1,
+        /* .y_chroma_shift = */ 1,
+        /* .depth          = */ 8,
+      },
+  /* [PIX_FMT_NV21] = */ {
+        /* .format         = */ PIX_FMT_NV21,
+        /* .name           = */ "nv21",
+        /* .nb_channels    = */ 2,
+        /* .color_type     = */ FF_COLOR_YUV,
+        /* .pixel_type     = */ FF_PIXEL_PACKED,
+        /* .is_alpha       = */ 0,
+        /* .x_chroma_shift = */ 1,
+        /* .y_chroma_shift = */ 1,
+        /* .depth          = */ 8,
+      },
   /* [PIX_FMT_YUV422P] = */ {
         /* .format         = */ PIX_FMT_YUV422P,
         /* .name           = */ "yuv422p",
@@ -1082,6 +1104,53 @@ yuv420p_to_yuv422 (AVPicture * dst, const AVPicture * src,
   }
 }
 
+static void
+nv12_to_nv21 (AVPicture * dst, const AVPicture * src, int width, int height)
+{
+  const uint8_t *s_c_ptr;
+  uint8_t *d_c_ptr;
+  int w;
+
+  s_c_ptr = src->data[1];
+  d_c_ptr = dst->data[1];
+
+  for (; height >= 2; height -= 2) {
+    for (w = width; w >= 2; w -= 2) {
+      d_c_ptr[0] = s_c_ptr[1];
+      d_c_ptr[1] = s_c_ptr[0];
+      s_c_ptr += 2;
+      d_c_ptr += 2;
+    }
+
+    /* handle odd width */
+    if (w) {
+      d_c_ptr[0] = s_c_ptr[1];
+      d_c_ptr[1] = s_c_ptr[0];
+      s_c_ptr += 2;
+      d_c_ptr += 2;
+    }
+  }
+  /* handle odd height */
+  if (height) {
+    for (w = width; w >= 2; w -= 2) {
+      d_c_ptr[0] = s_c_ptr[1];
+      d_c_ptr[1] = s_c_ptr[0];
+      s_c_ptr += 2;
+      d_c_ptr += 2;
+    }
+
+    /* handle odd width */
+    if (w) {
+      d_c_ptr[0] = s_c_ptr[1];
+      d_c_ptr[1] = s_c_ptr[0];
+      s_c_ptr += 2;
+      d_c_ptr += 2;
+    }
+  }
+}
+
+#define nv21_to_nv12 nv12_to_nv21
+
 #define SCALEBITS 10
 #define ONE_HALF  (1 << (SCALEBITS - 1))
 #define FIX(x)    ((int) ((x) * (1<<SCALEBITS) + 0.5))
@@ -2037,6 +2106,34 @@ static ConvertEntry convert_table[] = {
   {PIX_FMT_YUV420P, PIX_FMT_ARGB32, yuv420p_to_argb32},
   {PIX_FMT_YUV420P, PIX_FMT_ABGR32, yuv420p_to_abgr32},
 
+  {PIX_FMT_NV12, PIX_FMT_RGB555, nv12_to_rgb555},
+  {PIX_FMT_NV12, PIX_FMT_RGB565, nv12_to_rgb565},
+  {PIX_FMT_NV12, PIX_FMT_BGR24, nv12_to_bgr24},
+  {PIX_FMT_NV12, PIX_FMT_RGB24, nv12_to_rgb24},
+  {PIX_FMT_NV12, PIX_FMT_RGB32, nv12_to_rgb32},
+  {PIX_FMT_NV12, PIX_FMT_BGR32, nv12_to_bgr32},
+  {PIX_FMT_NV12, PIX_FMT_xRGB32, nv12_to_xrgb32},
+  {PIX_FMT_NV12, PIX_FMT_BGRx32, nv12_to_bgrx32},
+  {PIX_FMT_NV12, PIX_FMT_RGBA32, nv12_to_rgba32},
+  {PIX_FMT_NV12, PIX_FMT_BGRA32, nv12_to_bgra32},
+  {PIX_FMT_NV12, PIX_FMT_ARGB32, nv12_to_argb32},
+  {PIX_FMT_NV12, PIX_FMT_ABGR32, nv12_to_abgr32},
+  {PIX_FMT_NV12, PIX_FMT_ABGR32, nv12_to_nv21},
+
+  {PIX_FMT_NV21, PIX_FMT_RGB555, nv21_to_rgb555},
+  {PIX_FMT_NV21, PIX_FMT_RGB565, nv21_to_rgb565},
+  {PIX_FMT_NV21, PIX_FMT_BGR24, nv21_to_bgr24},
+  {PIX_FMT_NV21, PIX_FMT_RGB24, nv21_to_rgb24},
+  {PIX_FMT_NV21, PIX_FMT_RGB32, nv21_to_rgb32},
+  {PIX_FMT_NV21, PIX_FMT_BGR32, nv21_to_bgr32},
+  {PIX_FMT_NV21, PIX_FMT_xRGB32, nv21_to_xrgb32},
+  {PIX_FMT_NV21, PIX_FMT_BGRx32, nv21_to_bgrx32},
+  {PIX_FMT_NV21, PIX_FMT_RGBA32, nv21_to_rgba32},
+  {PIX_FMT_NV21, PIX_FMT_BGRA32, nv21_to_bgra32},
+  {PIX_FMT_NV21, PIX_FMT_ARGB32, nv21_to_argb32},
+  {PIX_FMT_NV21, PIX_FMT_ABGR32, nv21_to_abgr32},
+  {PIX_FMT_NV12, PIX_FMT_ABGR32, nv21_to_nv12},
+
   {PIX_FMT_YUV422P, PIX_FMT_YUV422, yuv422p_to_yuv422},
   {PIX_FMT_YUV422P, PIX_FMT_UYVY422, yuv422p_to_uyvy422},
 
@@ -2064,6 +2161,8 @@ static ConvertEntry convert_table[] = {
   {PIX_FMT_UYVY422, PIX_FMT_YUV422P, uyvy422_to_yuv422p},
 
   {PIX_FMT_RGB24, PIX_FMT_YUV420P, rgb24_to_yuv420p},
+  {PIX_FMT_RGB24, PIX_FMT_NV12, rgb24_to_nv12},
+  {PIX_FMT_RGB24, PIX_FMT_NV21, rgb24_to_nv21},
   {PIX_FMT_RGB24, PIX_FMT_RGB565, rgb24_to_rgb565},
   {PIX_FMT_RGB24, PIX_FMT_RGB555, rgb24_to_rgb555},
   {PIX_FMT_RGB24, PIX_FMT_RGB32, rgb24_to_rgb32},
@@ -2086,11 +2185,15 @@ static ConvertEntry convert_table[] = {
   {PIX_FMT_RGB32, PIX_FMT_RGB555, rgba32_to_rgb555},
   {PIX_FMT_RGB32, PIX_FMT_PAL8, rgb32_to_pal8},
   {PIX_FMT_RGB32, PIX_FMT_YUV420P, rgb32_to_yuv420p},
+  {PIX_FMT_RGB32, PIX_FMT_NV12, rgb32_to_nv12},
+  {PIX_FMT_RGB32, PIX_FMT_NV21, rgb32_to_nv21},
   {PIX_FMT_RGB32, PIX_FMT_GRAY8, rgb32_to_gray},
 
   {PIX_FMT_xRGB32, PIX_FMT_RGB24, xrgb32_to_rgb24},
   {PIX_FMT_xRGB32, PIX_FMT_PAL8, xrgb32_to_pal8},
   {PIX_FMT_xRGB32, PIX_FMT_YUV420P, xrgb32_to_yuv420p},
+  {PIX_FMT_xRGB32, PIX_FMT_NV12, xrgb32_to_nv12},
+  {PIX_FMT_xRGB32, PIX_FMT_NV21, xrgb32_to_nv21},
   {PIX_FMT_xRGB32, PIX_FMT_GRAY8, xrgb32_to_gray},
 
   {PIX_FMT_RGBA32, PIX_FMT_BGRA32, rgba32_to_bgra32},
@@ -2103,47 +2206,65 @@ static ConvertEntry convert_table[] = {
   {PIX_FMT_RGBA32, PIX_FMT_RGB555, rgba32_to_rgb555},
   {PIX_FMT_RGBA32, PIX_FMT_PAL8, rgba32_to_pal8},
   {PIX_FMT_RGBA32, PIX_FMT_YUV420P, rgba32_to_yuv420p},
+  {PIX_FMT_RGBA32, PIX_FMT_NV12, rgba32_to_nv12},
+  {PIX_FMT_RGBA32, PIX_FMT_NV21, rgba32_to_nv21},
   {PIX_FMT_RGBA32, PIX_FMT_GRAY8, rgba32_to_gray},
   {PIX_FMT_RGBA32, PIX_FMT_AYUV4444, rgba32_to_ayuv4444},
 
   {PIX_FMT_BGR24, PIX_FMT_RGB24, bgr24_to_rgb24},
   {PIX_FMT_BGR24, PIX_FMT_YUV420P, bgr24_to_yuv420p},
+  {PIX_FMT_BGR24, PIX_FMT_NV12, bgr24_to_nv12},
+  {PIX_FMT_BGR24, PIX_FMT_NV21, bgr24_to_nv21},
   {PIX_FMT_BGR24, PIX_FMT_GRAY8, bgr24_to_gray},
 
   {PIX_FMT_BGR32, PIX_FMT_RGB24, bgr32_to_rgb24},
   {PIX_FMT_BGR32, PIX_FMT_RGBA32, bgr32_to_rgba32},
   {PIX_FMT_BGR32, PIX_FMT_YUV420P, bgr32_to_yuv420p},
+  {PIX_FMT_BGR32, PIX_FMT_NV12, bgr32_to_nv12},
+  {PIX_FMT_BGR32, PIX_FMT_NV21, bgr32_to_nv21},
   {PIX_FMT_BGR32, PIX_FMT_GRAY8, bgr32_to_gray},
 
   {PIX_FMT_BGRx32, PIX_FMT_RGB24, bgrx32_to_rgb24},
   {PIX_FMT_BGRx32, PIX_FMT_RGBA32, bgrx32_to_rgba32},
   {PIX_FMT_BGRx32, PIX_FMT_YUV420P, bgrx32_to_yuv420p},
+  {PIX_FMT_BGRx32, PIX_FMT_NV12, bgrx32_to_nv12},
+  {PIX_FMT_BGRx32, PIX_FMT_NV21, bgrx32_to_nv21},
   {PIX_FMT_BGRx32, PIX_FMT_GRAY8, bgrx32_to_gray},
 
   {PIX_FMT_BGRA32, PIX_FMT_RGB24, bgra32_to_rgb24},
   {PIX_FMT_BGRA32, PIX_FMT_RGBA32, bgra32_to_rgba32},
   {PIX_FMT_BGRA32, PIX_FMT_YUV420P, bgra32_to_yuv420p},
+  {PIX_FMT_BGRA32, PIX_FMT_NV12, bgra32_to_nv12},
+  {PIX_FMT_BGRA32, PIX_FMT_NV21, bgra32_to_nv21},
   {PIX_FMT_BGRA32, PIX_FMT_GRAY8, bgra32_to_gray},
   {PIX_FMT_BGRA32, PIX_FMT_AYUV4444, bgra32_to_ayuv4444},
 
   {PIX_FMT_ABGR32, PIX_FMT_RGB24, abgr32_to_rgb24},
   {PIX_FMT_ABGR32, PIX_FMT_RGBA32, abgr32_to_rgba32},
   {PIX_FMT_ABGR32, PIX_FMT_YUV420P, abgr32_to_yuv420p},
+  {PIX_FMT_ABGR32, PIX_FMT_NV12, abgr32_to_nv12},
+  {PIX_FMT_ABGR32, PIX_FMT_NV21, abgr32_to_nv21},
   {PIX_FMT_ABGR32, PIX_FMT_GRAY8, abgr32_to_gray},
 
   {PIX_FMT_ARGB32, PIX_FMT_RGB24, argb32_to_rgb24},
   {PIX_FMT_ARGB32, PIX_FMT_RGBA32, argb32_to_rgba32},
   {PIX_FMT_ARGB32, PIX_FMT_YUV420P, argb32_to_yuv420p},
+  {PIX_FMT_ARGB32, PIX_FMT_NV12, argb32_to_nv12},
+  {PIX_FMT_ARGB32, PIX_FMT_NV21, argb32_to_nv21},
   {PIX_FMT_ARGB32, PIX_FMT_GRAY8, argb32_to_gray},
 
   {PIX_FMT_RGB555, PIX_FMT_RGB24, rgb555_to_rgb24},
   {PIX_FMT_RGB555, PIX_FMT_RGB32, rgb555_to_rgba32},
   {PIX_FMT_RGB555, PIX_FMT_RGBA32, rgb555_to_rgba32},
   {PIX_FMT_RGB555, PIX_FMT_YUV420P, rgb555_to_yuv420p},
+  {PIX_FMT_RGB555, PIX_FMT_NV12, rgb555_to_nv12},
+  {PIX_FMT_RGB555, PIX_FMT_NV21, rgb555_to_nv21},
   {PIX_FMT_RGB555, PIX_FMT_GRAY8, rgb555_to_gray},
 
   {PIX_FMT_RGB565, PIX_FMT_RGB24, rgb565_to_rgb24},
   {PIX_FMT_RGB565, PIX_FMT_YUV420P, rgb565_to_yuv420p},
+  {PIX_FMT_RGB565, PIX_FMT_NV12, rgb565_to_nv12},
+  {PIX_FMT_RGB565, PIX_FMT_NV21, rgb565_to_nv21},
   {PIX_FMT_RGB565, PIX_FMT_GRAY8, rgb565_to_gray},
 
   {PIX_FMT_GRAY8, PIX_FMT_RGB555, gray_to_rgb555},

gst/ffmpegcolorspace/imgconvert_template.h

@@ -103,7 +103,7 @@ static void glue(yuv420p_to_, RGB_NAME)(AVPicture *dst, const AVPicture *src,
         /* handle width */
         if (w) {
             YUV_TO_RGB1_CCIR(cb_ptr[0], cr_ptr[0]);
-            /* output 2 pixels */
+            /* output 2 pixel */
             YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
             RGB_OUT(d1, r, g, b);
             d1 += BPP;
@@ -115,6 +115,180 @@ static void glue(yuv420p_to_, RGB_NAME)(AVPicture *dst, const AVPicture *src,
     }
 }
 
+static void glue(nv12_to_, RGB_NAME)(AVPicture *dst, const AVPicture *src,
+                                        int width, int height)
+{
+    const uint8_t *y1_ptr, *y2_ptr, *c_ptr;
+    uint8_t *d, *d1, *d2;
+    int w, y, cb, cr, r_add, g_add, b_add;
+    uint8_t *cm = cropTbl + MAX_NEG_CROP;
+    unsigned int r, g, b;
+
+    d = dst->data[0];
+    y1_ptr = src->data[0];
+    c_ptr = src->data[1];
+    for(;height >= 2; height -= 2) {
+        d1 = d;
+        d2 = d + dst->linesize[0];
+        y2_ptr = y1_ptr + src->linesize[0];
+        for(w = width; w >= 2; w -= 2) {
+            YUV_TO_RGB1_CCIR(c_ptr[0], c_ptr[1]);
+            /* output 4 pixels */
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[1]);
+            RGB_OUT(d1 + BPP, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y2_ptr[0]);
+            RGB_OUT(d2, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y2_ptr[1]);
+            RGB_OUT(d2 + BPP, r, g, b);
+
+            d1 += 2 * BPP;
+            d2 += 2 * BPP;
+
+            y1_ptr += 2;
+            y2_ptr += 2;
+            c_ptr += 2;
+        }
+        /* handle odd width */
+        if (w) {
+            YUV_TO_RGB1_CCIR(c_ptr[0], c_ptr[1]);
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y2_ptr[0]);
+            RGB_OUT(d2, r, g, b);
+            d1 += BPP;
+            d2 += BPP;
+            y1_ptr++;
+            y2_ptr++;
+            c_ptr += 2;
+        }
+        d += 2 * dst->linesize[0];
+        y1_ptr += 2 * src->linesize[0] - width;
+        c_ptr += src->linesize[1] - width;
+    }
+    /* handle odd height */
+    if (height) {
+        d1 = d;
+        for(w = width; w >= 2; w -= 2) {
+            YUV_TO_RGB1_CCIR(c_ptr[0], c_ptr[1]);
+            /* output 2 pixels */
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[1]);
+            RGB_OUT(d1 + BPP, r, g, b);
+
+            d1 += 2 * BPP;
+
+            y1_ptr += 2;
+            c_ptr += 2;
+        }
+        /* handle odd width */
+        if (w) {
+            YUV_TO_RGB1_CCIR(c_ptr[0], c_ptr[1]);
+            /* output 1 pixel */
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+            d1 += BPP;
+
+            y1_ptr++;
+            c_ptr += 2;
+        }
+    }
+}
+
+static void glue(nv21_to_, RGB_NAME)(AVPicture *dst, const AVPicture *src,
+                                        int width, int height)
+{
+    const uint8_t *y1_ptr, *y2_ptr, *c_ptr;
+    uint8_t *d, *d1, *d2;
+    int w, y, cb, cr, r_add, g_add, b_add;
+    uint8_t *cm = cropTbl + MAX_NEG_CROP;
+    unsigned int r, g, b;
+
+    d = dst->data[0];
+    y1_ptr = src->data[0];
+    c_ptr = src->data[1];
+    for(;height >= 2; height -= 2) {
+        d1 = d;
+        d2 = d + dst->linesize[0];
+        y2_ptr = y1_ptr + src->linesize[0];
+        for(w = width; w >= 2; w -= 2) {
+            YUV_TO_RGB1_CCIR(c_ptr[1], c_ptr[0]);
+            /* output 4 pixels */
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[1]);
+            RGB_OUT(d1 + BPP, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y2_ptr[0]);
+            RGB_OUT(d2, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y2_ptr[1]);
+            RGB_OUT(d2 + BPP, r, g, b);
+
+            d1 += 2 * BPP;
+            d2 += 2 * BPP;
+
+            y1_ptr += 2;
+            y2_ptr += 2;
+            c_ptr += 2;
+        }
+        /* handle odd width */
+        if (w) {
+            YUV_TO_RGB1_CCIR(c_ptr[1], c_ptr[0]);
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y2_ptr[0]);
+            RGB_OUT(d2, r, g, b);
+            d1 += BPP;
+            d2 += BPP;
+            y1_ptr++;
+            y2_ptr++;
+            c_ptr += 2;
+        }
+        d += 2 * dst->linesize[0];
+        y1_ptr += 2 * src->linesize[0] - width;
+        c_ptr += src->linesize[1] - width;
+    }
+    /* handle odd height */
+    if (height) {
+        d1 = d;
+        for(w = width; w >= 2; w -= 2) {
+            YUV_TO_RGB1_CCIR(c_ptr[1], c_ptr[0]);
+            /* output 2 pixels */
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[1]);
+            RGB_OUT(d1 + BPP, r, g, b);
+
+            d1 += 2 * BPP;
+
+            y1_ptr += 2;
+            c_ptr += 2;
+        }
+        /* handle odd width */
+        if (w) {
+            YUV_TO_RGB1_CCIR(c_ptr[1], c_ptr[0]);
+            /* output 1 pixel */
+            YUV_TO_RGB2_CCIR(r, g, b, y1_ptr[0]);
+            RGB_OUT(d1, r, g, b);
+            d1 += BPP;
+
+            y1_ptr++;
+            c_ptr += 2;
+        }
+    }
+}
+
 static void glue(yuvj420p_to_, RGB_NAME)(AVPicture *dst, const AVPicture *src,
                                          int width, int height)
 {
@@ -317,6 +491,214 @@ static void glue(RGB_NAME, _to_yuv420p)(AVPicture *dst, const AVPicture *src,
     }
 }
 
+static void glue(RGB_NAME, _to_nv12)(AVPicture *dst, const AVPicture *src,
+                                        int width, int height)
+{
+    int wrap, wrap3;
+    int r, g, b, r1, g1, b1, w;
+    uint8_t *lum, *c;
+    const uint8_t *p;
+
+    lum = dst->data[0];
+    c = dst->data[1];
+
+    wrap = dst->linesize[0];
+    wrap3 = src->linesize[0];
+    p = src->data[0];
+    for(;height>=2;height -= 2) {
+        for(w = width; w >= 2; w -= 2) {
+            RGB_IN(r, g, b, p);
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+
+            RGB_IN(r, g, b, p + BPP);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = RGB_TO_Y_CCIR(r, g, b);
+            p += wrap3;
+            lum += wrap;
+
+            RGB_IN(r, g, b, p);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+
+            RGB_IN(r, g, b, p + BPP);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = RGB_TO_Y_CCIR(r, g, b);
+
+            c[0] = RGB_TO_U_CCIR(r1, g1, b1, 2);
+            c[1] = RGB_TO_V_CCIR(r1, g1, b1, 2);
+
+
+            c += 2;
+            p += -wrap3 + 2 * BPP;
+            lum += -wrap + 2;
+        }
+	/* handle odd width */
+        if (w) {
+            RGB_IN(r, g, b, p);
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+            p += wrap3;
+            lum += wrap;
+            RGB_IN(r, g, b, p);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+            c[0] = RGB_TO_U_CCIR(r1, g1, b1, 1);
+            c[1] = RGB_TO_V_CCIR(r1, g1, b1, 1);
+            c += 2;
+            p += -wrap3 + BPP;
+            lum += -wrap + 1;
+        }
+        p += wrap3 + (wrap3 - width * BPP);
+        lum += wrap + (wrap - width);
+        c += dst->linesize[1] - width;
+    }
+    /* handle odd height */
+    if (height) {
+        for(w = width; w >= 2; w -= 2) {
+            RGB_IN(r, g, b, p);
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+
+            RGB_IN(r, g, b, p + BPP);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = RGB_TO_Y_CCIR(r, g, b);
+            c[0] = RGB_TO_U_CCIR(r1, g1, b1, 1);
+            c[1] = RGB_TO_V_CCIR(r1, g1, b1, 1);
+            c += 2;
+            p += 2 * BPP;
+            lum += 2;
+        }
+	/* handle odd width */
+        if (w) {
+            RGB_IN(r, g, b, p);
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+            c[0] = RGB_TO_U_CCIR(r, g, b, 0);
+            c[1] = RGB_TO_V_CCIR(r, g, b, 0);
+        }
+    }
+}
+
+static void glue(RGB_NAME, _to_nv21)(AVPicture *dst, const AVPicture *src,
+                                        int width, int height)
+{
+    int wrap, wrap3;
+    int r, g, b, r1, g1, b1, w;
+    uint8_t *lum, *c;
+    const uint8_t *p;
+
+    lum = dst->data[0];
+    c = dst->data[1];
+
+    wrap = dst->linesize[0];
+    wrap3 = src->linesize[0];
+    p = src->data[0];
+    for(;height>=2;height -= 2) {
+        for(w = width; w >= 2; w -= 2) {
+            RGB_IN(r, g, b, p);
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+
+            RGB_IN(r, g, b, p + BPP);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = RGB_TO_Y_CCIR(r, g, b);
+            p += wrap3;
+            lum += wrap;
+
+            RGB_IN(r, g, b, p);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+
+            RGB_IN(r, g, b, p + BPP);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = RGB_TO_Y_CCIR(r, g, b);
+
+            c[1] = RGB_TO_U_CCIR(r1, g1, b1, 2);
+            c[0] = RGB_TO_V_CCIR(r1, g1, b1, 2);
+
+
+            c += 2;
+            p += -wrap3 + 2 * BPP;
+            lum += -wrap + 2;
+        }
+	/* handle odd width */
+        if (w) {
+            RGB_IN(r, g, b, p);
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+            p += wrap3;
+            lum += wrap;
+            RGB_IN(r, g, b, p);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+            c[1] = RGB_TO_U_CCIR(r1, g1, b1, 1);
+            c[0] = RGB_TO_V_CCIR(r1, g1, b1, 1);
+            c += 2;
+            p += -wrap3 + BPP;
+            lum += -wrap + 1;
+        }
+        p += wrap3 + (wrap3 - width * BPP);
+        lum += wrap + (wrap - width);
+        c += dst->linesize[1] - width;
+    }
+    /* handle odd height */
+    if (height) {
+        for(w = width; w >= 2; w -= 2) {
+            RGB_IN(r, g, b, p);
+            r1 = r;
+            g1 = g;
+            b1 = b;
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+
+            RGB_IN(r, g, b, p + BPP);
+            r1 += r;
+            g1 += g;
+            b1 += b;
+            lum[1] = RGB_TO_Y_CCIR(r, g, b);
+            c[1] = RGB_TO_U_CCIR(r1, g1, b1, 1);
+            c[0] = RGB_TO_V_CCIR(r1, g1, b1, 1);
+            c += 2;
+            p += 2 * BPP;
+            lum += 2;
+        }
+	/* handle odd width */
+        if (w) {
+            RGB_IN(r, g, b, p);
+            lum[0] = RGB_TO_Y_CCIR(r, g, b);
+            c[1] = RGB_TO_U_CCIR(r, g, b, 0);
+            c[0] = RGB_TO_V_CCIR(r, g, b, 0);
+        }
+    }
+}
+
 static void glue(RGB_NAME, _to_gray)(AVPicture *dst, const AVPicture *src,
                                      int width, int height)
 {