Arbitrary-sized image compression without padding

摘要

The performance of image compression is based on number of levels of wavelet transform and compression algorithm. Padding some extra pixels to the inappropriate-sized images can make them legible for many levels of conventional (or symmetric) wavelet transform. However, more pixels mean more cost in bit budget and also more computational time. The implementation of asymmetric wavelet decomposition and quadtree prediction scheme for arbitrary-size image are proposed in this paper. The main encoding algorithm used is Context Adaptive Wavelet Different Reduction (CAWDR) without list, which requires a small number of memory during the encoding process and has fast algorithm. The non-padding method yields comparable results to both the padding one and the other state-of-the-art coders in the literature.