Y4M input support for 4:2:2, 4:4:4, 4:4:4:4

Adds a new experiment CONFIG_NON420 that allows other chroma subsamplings
to be passed to the codec. This commit allows the data to be passed from
a y4m input file through vpxenc to the codec, where they're currently
rejected. Later commits will finish support for this inside the codec.

Change-Id: Ib3aac604d8cad9e24cef395fa1067f16ba7e8e43

diff --git a/configure b/configure
index 6e60af1..573ec36 100755 (executable)
--- a/configure
+++ b/configure
@@ -250,6 +250,7 @@ EXPERIMENT_LIST="
     multiple_arf
     code_zerogroup
     sb8x8
+    non420
 "
 CONFIG_LIST="
     external_build

diff --git a/vpx/vpx_image.h b/vpx/vpx_image.h
index 809fa38..c304bac 100644 (file)
--- a/vpx/vpx_image.h
+++ b/vpx/vpx_image.h
@@ -55,9 +55,11 @@ extern "C" {
     VPX_IMG_FMT_YV12    = VPX_IMG_FMT_PLANAR | VPX_IMG_FMT_UV_FLIP | 1, /**< planar YVU */
     VPX_IMG_FMT_I420    = VPX_IMG_FMT_PLANAR | 2,
     VPX_IMG_FMT_VPXYV12 = VPX_IMG_FMT_PLANAR | VPX_IMG_FMT_UV_FLIP | 3, /** < planar 4:2:0 format with vpx color space */
-    VPX_IMG_FMT_VPXI420 = VPX_IMG_FMT_PLANAR | 4   /** < planar 4:2:0 format with vpx color space */
-  }
-                        vpx_img_fmt_t; /**< alias for enum vpx_img_fmt */
+    VPX_IMG_FMT_VPXI420 = VPX_IMG_FMT_PLANAR | 4,
+    VPX_IMG_FMT_I422    = VPX_IMG_FMT_PLANAR | 5,
+    VPX_IMG_FMT_I444    = VPX_IMG_FMT_PLANAR | 6,
+    VPX_IMG_FMT_444A    = VPX_IMG_FMT_PLANAR | VPX_IMG_FMT_HAS_ALPHA | 7
+  } vpx_img_fmt_t; /**< alias for enum vpx_img_fmt */
 
 #if !defined(VPX_CODEC_DISABLE_COMPAT) || !VPX_CODEC_DISABLE_COMPAT
 #define IMG_FMT_PLANAR         VPX_IMG_FMT_PLANAR     /**< \deprecated Use #VPX_IMG_FMT_PLANAR */

diff --git a/vpxenc.c b/vpxenc.c
index 95c6cf2..33a56a4 100644 (file)
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -326,6 +326,7 @@ struct input_state {
   unsigned int          h;
   struct vpx_rational   framerate;
   int                   use_i420;
+  int                   only_i420;
 };
 
 
@@ -1793,7 +1794,8 @@ void open_input_file(struct input_state *input) {
 
   if (input->detect.buf_read == 4
       && file_is_y4m(input->file, &input->y4m, input->detect.buf)) {
-    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4) >= 0) {
+    if (y4m_input_open(&input->y4m, input->file, input->detect.buf, 4,
+                       input->only_i420) >= 0) {
       input->file_type = FILE_TYPE_Y4M;
       input->w = input->y4m.pic_w;
       input->h = input->y4m.pic_h;
@@ -2517,6 +2519,7 @@ int main(int argc, const char **argv_) {
   input.framerate.num = 30;
   input.framerate.den = 1;
   input.use_i420 = 1;
+  input.only_i420 = 1;
 
   /* First parse the global configuration values, because we want to apply
    * other parameters on top of the default configuration provided by the
@@ -2551,6 +2554,12 @@ int main(int argc, const char **argv_) {
   if (!input.fn)
     usage_exit();
 
+#if CONFIG_NON420
+  /* Decide if other chroma subsamplings than 4:2:0 are supported */
+  if (global.codec->fourcc == VP9_FOURCC)
+    input.only_i420 = 0;
+#endif
+
   for (pass = global.pass ? global.pass - 1 : 0; pass < global.passes; pass++) {
     int frames_in = 0, seen_frames = 0;
     int64_t estimated_time_left = -1;

diff --git a/y4minput.c b/y4minput.c
index 24f0c15..47f005a 100644 (file)
--- a/y4minput.c
+++ b/y4minput.c
@@ -659,7 +659,8 @@ static void y4m_convert_null(y4m_input *_y4m, unsigned char *_dst,
                              unsigned char *_aux) {
 }
 
-int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip) {
+int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip,
+                   int only_420) {
   char buffer[80];
   int  ret;
   int  i;
@@ -701,6 +702,8 @@ int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip) {
             "Only progressive scan handled.\n");
     return -1;
   }
+  _y4m->vpx_fmt = VPX_IMG_FMT_I420;
+  _y4m->vpx_bps = 12;
   if (strcmp(_y4m->chroma_type, "420") == 0 ||
       strcmp(_y4m->chroma_type, "420jpeg") == 0) {
     _y4m->src_c_dec_h = _y4m->dst_c_dec_h = _y4m->src_c_dec_v = _y4m->dst_c_dec_v = 2;
@@ -734,16 +737,30 @@ int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip) {
     _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
     _y4m->convert = y4m_convert_422jpeg_420jpeg;
   } else if (strcmp(_y4m->chroma_type, "422") == 0) {
-    _y4m->src_c_dec_h = _y4m->dst_c_dec_h = 2;
+    _y4m->src_c_dec_h = 2;
     _y4m->src_c_dec_v = 1;
-    _y4m->dst_c_dec_v = 2;
-    _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h;
-    /*Chroma filter required: read into the aux buf first.
-      We need to make two filter passes, so we need some extra space in the
-       aux buffer.*/
-    _y4m->aux_buf_read_sz = 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
-    _y4m->aux_buf_sz = _y4m->aux_buf_read_sz + ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
-    _y4m->convert = y4m_convert_422_420jpeg;
+    if (only_420) {
+      _y4m->dst_c_dec_h = 2;
+      _y4m->dst_c_dec_v = 2;
+      _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h;
+      /*Chroma filter required: read into the aux buf first.
+        We need to make two filter passes, so we need some extra space in the
+         aux buffer.*/
+      _y4m->aux_buf_read_sz = 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
+      _y4m->aux_buf_sz = _y4m->aux_buf_read_sz +
+          ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
+      _y4m->convert = y4m_convert_422_420jpeg;
+    } else {
+      _y4m->vpx_fmt = VPX_IMG_FMT_I422;
+      _y4m->vpx_bps = 16;
+      _y4m->dst_c_dec_h = _y4m->src_c_dec_h;
+      _y4m->dst_c_dec_v = _y4m->src_c_dec_v;
+      _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h
+                              + 2 * ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
+      /*Natively supported: no conversion required.*/
+      _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0;
+      _y4m->convert = y4m_convert_null;
+      }
   } else if (strcmp(_y4m->chroma_type, "411") == 0) {
     _y4m->src_c_dec_h = 4;
     _y4m->dst_c_dec_h = 2;
@@ -758,29 +775,52 @@ int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip) {
     _y4m->convert = y4m_convert_411_420jpeg;
   } else if (strcmp(_y4m->chroma_type, "444") == 0) {
     _y4m->src_c_dec_h = 1;
-    _y4m->dst_c_dec_h = 2;
     _y4m->src_c_dec_v = 1;
-    _y4m->dst_c_dec_v = 2;
-    _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h;
-    /*Chroma filter required: read into the aux buf first.
-      We need to make two filter passes, so we need some extra space in the
-       aux buffer.*/
-    _y4m->aux_buf_read_sz = 2 * _y4m->pic_w * _y4m->pic_h;
-    _y4m->aux_buf_sz = _y4m->aux_buf_read_sz + ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
-    _y4m->convert = y4m_convert_444_420jpeg;
+    if (only_420) {
+      _y4m->dst_c_dec_h = 2;
+      _y4m->dst_c_dec_v = 2;
+      _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h;
+      /*Chroma filter required: read into the aux buf first.
+        We need to make two filter passes, so we need some extra space in the
+         aux buffer.*/
+      _y4m->aux_buf_read_sz = 2 * _y4m->pic_w * _y4m->pic_h;
+      _y4m->aux_buf_sz = _y4m->aux_buf_read_sz +
+          ((_y4m->pic_w + 1) / 2) * _y4m->pic_h;
+      _y4m->convert = y4m_convert_444_420jpeg;
+    } else {
+      _y4m->vpx_fmt = VPX_IMG_FMT_I444;
+      _y4m->vpx_bps = 24;
+      _y4m->dst_c_dec_h = _y4m->src_c_dec_h;
+      _y4m->dst_c_dec_v = _y4m->src_c_dec_v;
+      _y4m->dst_buf_read_sz = 3 * _y4m->pic_w * _y4m->pic_h;
+      /*Natively supported: no conversion required.*/
+      _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0;
+      _y4m->convert = y4m_convert_null;
+    }
   } else if (strcmp(_y4m->chroma_type, "444alpha") == 0) {
     _y4m->src_c_dec_h = 1;
-    _y4m->dst_c_dec_h = 2;
     _y4m->src_c_dec_v = 1;
-    _y4m->dst_c_dec_v = 2;
-    _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h;
-    /*Chroma filter required: read into the aux buf first.
-      We need to make two filter passes, so we need some extra space in the
-       aux buffer.
-      The extra plane also gets read into the aux buf.
-      It will be discarded.*/
-    _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 3 * _y4m->pic_w * _y4m->pic_h;
-    _y4m->convert = y4m_convert_444_420jpeg;
+    if (only_420) {
+      _y4m->dst_c_dec_h = 2;
+      _y4m->dst_c_dec_v = 2;
+      _y4m->dst_buf_read_sz = _y4m->pic_w * _y4m->pic_h;
+      /*Chroma filter required: read into the aux buf first.
+        We need to make two filter passes, so we need some extra space in the
+         aux buffer.
+        The extra plane also gets read into the aux buf.
+        It will be discarded.*/
+      _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 3 * _y4m->pic_w * _y4m->pic_h;
+      _y4m->convert = y4m_convert_444_420jpeg;
+    } else {
+      _y4m->vpx_fmt = VPX_IMG_FMT_444A;
+      _y4m->vpx_bps = 32;
+      _y4m->dst_c_dec_h = _y4m->src_c_dec_h;
+      _y4m->dst_c_dec_v = _y4m->src_c_dec_v;
+      _y4m->dst_buf_read_sz = 4 * _y4m->pic_w * _y4m->pic_h;
+      /*Natively supported: no conversion required.*/
+      _y4m->aux_buf_sz = _y4m->aux_buf_read_sz = 0;
+      _y4m->convert = y4m_convert_null;
+    }
   } else if (strcmp(_y4m->chroma_type, "mono") == 0) {
     _y4m->src_c_dec_h = _y4m->src_c_dec_v = 0;
     _y4m->dst_c_dec_h = _y4m->dst_c_dec_v = 2;
@@ -847,22 +887,23 @@ int y4m_input_fetch_frame(y4m_input *_y4m, FILE *_fin, vpx_image_t *_img) {
      sizes, which would require a separate fread call for every row.*/
   memset(_img, 0, sizeof(*_img));
   /*Y4M has the planes in Y'CbCr order, which libvpx calls Y, U, and V.*/
-  _img->fmt = IMG_FMT_I420;
+  _img->fmt = _y4m->vpx_fmt;
   _img->w = _img->d_w = _y4m->pic_w;
   _img->h = _img->d_h = _y4m->pic_h;
-  /*This is hard-coded to 4:2:0 for now, as that's all VP8 supports.*/
-  _img->x_chroma_shift = 1;
-  _img->y_chroma_shift = 1;
-  _img->bps = 12;
+  _img->x_chroma_shift = _y4m->dst_c_dec_h >> 1;
+  _img->y_chroma_shift = _y4m->dst_c_dec_v >> 1;
+  _img->bps = _y4m->vpx_bps;
+
   /*Set up the buffer pointers.*/
   pic_sz = _y4m->pic_w * _y4m->pic_h;
   c_w = (_y4m->pic_w + _y4m->dst_c_dec_h - 1) / _y4m->dst_c_dec_h;
   c_h = (_y4m->pic_h + _y4m->dst_c_dec_v - 1) / _y4m->dst_c_dec_v;
   c_sz = c_w * c_h;
-  _img->stride[PLANE_Y] = _y4m->pic_w;
+  _img->stride[PLANE_Y] = _img->stride[PLANE_ALPHA] = _y4m->pic_w;
   _img->stride[PLANE_U] = _img->stride[PLANE_V] = c_w;
   _img->planes[PLANE_Y] = _y4m->dst_buf;
   _img->planes[PLANE_U] = _y4m->dst_buf + pic_sz;
   _img->planes[PLANE_V] = _y4m->dst_buf + pic_sz + c_sz;
+  _img->planes[PLANE_ALPHA] = _y4m->dst_buf + pic_sz + 2 * c_sz;
   return 1;
 }

diff --git a/y4minput.h b/y4minput.h
index 2fa3767..b2a390c 100644 (file)
--- a/y4minput.h
+++ b/y4minput.h
@@ -51,9 +51,12 @@ struct y4m_input {
   y4m_convert_func  convert;
   unsigned char    *dst_buf;
   unsigned char    *aux_buf;
+  enum vpx_img_fmt  vpx_fmt;
+  int               vpx_bps;
 };
 
-int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip);
+int y4m_input_open(y4m_input *_y4m, FILE *_fin, char *_skip, int _nskip,
+                   int only_420);
 void y4m_input_close(y4m_input *_y4m);
 int y4m_input_fetch_frame(y4m_input *_y4m, FILE *_fin, vpx_image_t *img);