Choice of Wavelets for Image Compression

In wavelet-based image coding the choice of wavelets is crucial and determines the coding

performance. Current techniques use computationally intensive search procedures to find the optimal basis (type,

order and tree). In this paper, we show that searching for optimal wavelet does not always offer a substantial

improvement in coding performance over "good" standard wavelets. We propose some guidelines for determining

the need to search for the "optimal" wavelets based on the statistics of the image to be coded. In addition, we

propose an adaptive wavelet packet decomposition algorithm based on the local transform gain of each stage of

the decomposition. The proposed algorithm provides a good coding performance at a substantially reduced

complexity.