In this work, a high-quality, space-frequency adaptive, subbandnimage codec is presented. The algorithm jointly optimizes space andnfrequency segmentation of an image. First, wavelet packets are formed tonlocalize the high-energy frequency bands. Then, the subband coefficientsnare further classified to maximize the coding gain. The design target isnthe minimization of Lagrangian cost functions: the cost is thendistortion in l2 norm and the constraint is the coding rate.nThe resultant mapping is used to quantize the subband coefficients withntrellis coded quantization. This is followed by adaptive arithmeticncoding producing the final compressed bitstream. The described approachnis tested on several images, and the results are compared to some otherncompression techniques
展开▼