Wavelet Filter Design for On-Board Image Compression

this paper, we report on tree-structured subband coder (or wavelet filter) designs which satisfy complexity constraints at the encoder. First, we review the theory of perfect reconstruction filter banks, and describe two families of solutions which are appropriate for our purposes. Then, we review definitions and known upper bounds on the subband coding gain of arbitrary linear transforms (not necessarily orthogonal). Finally, we derive solutions which maximize the subband coding gain, an incomplete, but informative measure of the performance of a subband transform-based coder, and give good integer approximations of these solutions. Additional remarks on the design of optimal boundary filters and the design of non-orthogonal integer block transforms are given