- // OpenGL has texel center in (0.5, 0.5), but different formats have
- // chroma in various other places. If luma samples are X, the chroma
- // sample is *, and subsampling is 3x3, the situation with chroma
- // center in (0.5, 0.5) looks approximately like this:
- //
- // X X
- // *
- // X X
- //
- // If, on the other hand, chroma center is in (0.0, 0.5) (common
- // for e.g. MPEG-4), the figure changes to:
- //
- // X X
- // *
- // X X
- //
- // In other words, (0.0, 0.0) means that the chroma sample is exactly
- // co-sited on top of the top-left luma sample. Note, however, that
- // this is _not_ 0.5 texels to the left, since the OpenGL's texel center
- // is in (0.5, 0.5); it is in (0.25, 0.25). In a sense, the four luma samples
- // define a square where chroma position (0.0, 0.0) is in texel position
- // (0.25, 0.25) and chroma position (1.0, 1.0) is in texel position (0.75, 0.75)
- // (the outer border shows the borders of the texel itself, ie. from
- // (0, 0) to (1, 1)):
- //
- // ---------
- // | |
- // | X---X |
- // | | * | |
- // | X---X |
- // | |
- // ---------
- //
- // Also note that if we have no subsampling, the square will have zero
- // area and the chroma position does not matter at all.
- float chroma_x_local_position =
- (0.5 + ycbcr_format.chroma_x_position * (ycbcr_format.chroma_subsampling_x - 1)) /
- ycbcr_format.chroma_subsampling_x;
- float chroma_y_local_position =
- (0.5 + ycbcr_format.chroma_y_position * (ycbcr_format.chroma_subsampling_y - 1)) /
- ycbcr_format.chroma_subsampling_y;