Context-modeled wavelet difference reduction coding based on fractional bit-plane partitioning

This paper presents a new embedded wavelet image compression algorithm that is based on the wavelet difference reduction (WDR) method developed by Tian and Wells. While WDR employs a fixed order on scanning of the coefficients in a wavelet pyramid, the proposed algorithm employs an adaptive scanning order that results from partitioning each bit-plane into multiple fractional bit-planes by exploiting intra-subband and inter-subband correlation. Arithmetic coding (AC) in WDR is avoided in the proposed coder, resulting in lower complexity. These approaches enable the proposed algorithm to be conceptually simple and outperform both WDR and SPIHT (without AC) by Said and Pearlman in a rate-distortion sense. The PSNR improvement over WDR ranges from 0.1 to 1.5 dB on the set of test images at various bitrates.