Embedded Morphological Dilation and Variable Run Length Wavelet Coding

Recent success in wavelet image compression is mainly attributed to reorganization and representation of wavelet coefficients. Morphological wavelet image coders are among state-of-the-art reorganization schemes and they represent the significance map by morphological dilation based on connectivity analysis to utilize the statistical property of clustering trend of most significant coefficients. A new reorganization scheme named embedded morphological dilation and variable run length wavelet coding is presented in this paper, which captures arbitrary cluster shape of significant coefficients and exploits statistical dependencies of zero-value coefficients within subband. Results show that the proposed algorithm by exploiting within-subband correlation outperforms embedded block coding with optimized truncation (EBCOT) and is competitive with the energy clustering and zero-quadtree representation (ECZQR).