Multistage Vector Quantization of Image by Using Clustering Interpolation in DCT Domain

Abstract

An adaptive image coding scheme, called multistage vector quantization (MSVQ) is proposed. A new algorithm of MSVQ uses clustering interpolation in DCT domain and enables vector quantization of large image blocks with tolerable encoding complexity. Their mean luminance values are efficiently removed by the spline interpolative vector quantizer using a small number of bits. It can achieve a very high compression rate for low detailed images without complicated classification of transform blocks and subvector construction.