Image compression based on optimal subband decomposition and vector quantization

A novel image coding scheme is presented which is based on optimal filter design for subband coding followed by vector quantization. The optimal filter design is based on minimization of Shannon's entropy which is considered as the cost function. The upper bound of the entropy of the decomposed image is obtained in terms of the filter response with the assumption that the p.d.f. of the pixel values of the image is Gaussian. The filter minimizes this upper bound of entropy in one of the subbands using steepest descent algorithm. The subband for which the entropy is minimized is discarded and the rest of the three subbands are coded using vector quantization with varying bit assignment for each subband. A significant compression ratio of 41.72 is achieved. Filter design and coding scheme is discussed and presented in detail in the paper along with the results.