Wavelet image coding using direction and power classified vector quantization

This paper proposes a new scheme for image coding using wavelet transforms. We treat a two-level transformed image with seven subimages. Apart from the lowest subimage, we classify the subimages into 3 groups by direction. In each group, we form 5-D crossband vectors, each consisting of one sample of level 2 subimage and 2/spl times/2 samples of level 1 subimage corresponding to the same spatial location in the original image. The resulted 3 vector groups are further classified into 3 classes by their vector powers, respectively. Then direction and power classified vectors are vector quantized using their own codebooks. The lowest subimage is coded by 2D-DPCM.