Fast and Efficient Basis Selection Methods for Embedded Wavelet Packet Image Coding

Abstract

Wavelet packet (WP) image coding algorithms have shown consistent improvement over those based on wavelet transform. However, most of the current work cannot produce an embedded bit stream, since an embedded WP image coding method requires a valid and uniform decomposition for a given image. In this paper, based on the idea of maximizing variance of each subband, we propose two rate-unrelated basis selection criteria for embedded image coding. According to these criteria, efficient decompositions are found by growing the decomposition tree once or by pruning the full decomposition tree once. The experimental results show that, with the same subband coding scheme, the proposed basis selection methods achieve better coding performance than that of previous selection criteria. Comparison with the rate-distortion optimized basis selection scheme shows that, the proposed methods have 0.33dB loss at most, with the advantages of an embedded coding fashion and lower computational complexity.