I. Introduction
In classical block-based color image compression, a source RGB image is usually converted to a YCbCr image to remove the redundancies across three color channels. Then, some necessary compression operations, including the discrete cosine transform (DCT), quantization, and entropy coding, are performed on the luminance component (Y) and two chrominance components (Cb and Cr) in the YCbCr space to perform the compression. Since most display devices only accept RGB signals as the inputs, the compressed YCbCr image needs to be converted back to the RGB space. Accordingly, the compression distortion should be measured in the RGB space to evaluate the quality of the compressed image. In practice, minimizing such a distortion requires a design of a high-efficient compression scheme for the RGB signals [1].