Image compression with full wavelet transform (FWT) and vector quantization

Image compression based on the full wavelet transform (FWT) was introduced recently by Wong and Kuo. This multiresolution approach recursively applies the two-scale wavelet decomposition to all sub-images so that an image is decomposed into blocks of the same size. One benefit of this approach is that energy compaction can be achieved in the spatial as well as frequency domains. Their paper describes a bit allocation and scalar quantization method that utilizes both forms of energy compaction and provides good results. Yet, it has been shown that vector quantization will perform better than scalar quantization when a suitable codebook is used. This paper describes some of the shortcomings of the bit allocation equation and suggests a method for compressing the wavelet coefficients using vector quantization. Experimental results using the method described are presented and reviewed