I. Introduction
Prior to compression, the input red-green-blue (RGB) full-color image is first transformed to a YUV image by (1), where (,,) and (,,) denote the original RGB triple value and the co-located YUV triple value, respectively, at pixel location , , in each RGB and YUV block-pair based on row-major order. Next, a chroma 4:2:0 subsampling scheme is adopted to subsample the one sampled U and V components, and , for each UV block. As shown in Fig. 1, the subsampled YUV image is then conveyed into the encoder for compression. Throughout this paper, we only discuss the chroma 4:2:0 subsampling, although our discussion is also applicable to chroma 4:2:2 subsampling. The commonly used seven chroma subsampling schemes are the 4:2:0(L), 4:2:0(R), 4:2:0(A), 4:2:0(DIRECT), 4:2:0(BRIGHT), 4:2:0(BRIGHT_MEAN) [13], and 4:2:0(MPEG-B) [16]. \begin{equation} \begin{bmatrix} Y_{i} \\ U_{i} \\ V_{i} \end{bmatrix} = \begin{bmatrix} 0.257 & 0.504 & 0.098 \\ -0.148 & -0.291 & 0.439 \\ 0.439 & -0.368 & -0.071 \end{bmatrix} \begin{bmatrix} R_{i}^{ori} \\ G_{i}^{ori} \\ B_{i}^{ori} \end{bmatrix} \!+\! \begin{bmatrix} 16 \\ 128 \\ 128 \end{bmatrix} \end{equation}
The conventional chroma subsampling scheme.