The framework for an image coding system based on embedded zerotrees consists of three stages: (i) wavelet transform (ii) embedded zerotree encoding and (iii) adaptive arithmetic encoding. In this framework, the selection of the wavelet filter becomes an important issue. In this paper, we present a modification to the scanning approach in the set partitioning algorithm proposed in Said and Pearlman (1996) to exploit the correlation in a local neighborhood. Two new criteria are proposed for evaluating the performance of wavelets in lossless image compression applications: zero tree count and monotone spectral ordering of subbands produced after the wavelet transform in a multiresolution scheme. We evaluate several wavelet filters to test the evaluation criteria and present experimental results to justify the proposed performance criteria.
展开▼