Wavelet image coding based on significance extraction using morphological operation

In this paper, an efficient embedded image compression algorithm is presented. It is based on the observation that the distribution of significant coefficients is intra-subband clustered and inter-subband self-similar. Morphological dilation is used to capture the clustered significant coefficients in each subband, resulting in the partitioning of each subband into significance clusters and insignificance space. Thus for entropy coding, different probability models are used for different regions according to their own probability distributions. When encoding the insignificant space, which contains mostly zeros, investigation reveals that the zerotree data structure is not very efficient to represent zeros across scales for texture images and a more efficient method is presented. Experimental results show that the proposed algorithm is very effective especially for texture images.