A hybrid image coding scheme using CVQ and DCT

Abstract

The authors describe a hybrid image coding scheme called the classified hybrid image coder (CHIC). CHIC combines the classified vector quantization (CVQ) and the discrete cosine transform (DCT) coding schemes to achieve a better performance than would be possible if they were to be used separately. The image is subdivided into 16 × 16 pixel blocks (vectors) and transformed using DCT. Each vector is classified into an edge vector and a shade vector. Classification is based on the distribution and value of the DCT coefficients in each vector. Shade vectors are coded using CVQ. Edge vectors are further classified according to their edge strengths. In each of these classes, the coefficients are quantized and then coded with an adaptive arithmetic coder (AAC). The performance of the AAC is further improved by further classifying the DCT coefficients into different groups, based on their different frequency ranges. The AAC uses a set of training images to generate the initial histogram tables.