Image coding using wavelets, interval wavelets and multi- layered wedgelets

Abstract

This work focuses on designing new bases and analysis techniques for approximating image edges. The first contribution is the proposal of the wavelet-based color set partitioning in hierarchical trees (CSPIHT) algorithm for embedded color image encoding, which is able to achieve comparable or better performance than other state-of-the-art techniques. In finite signal analysis, we design several new interval wavelets that allow interval wavelets to provide an efficient representation of image singularities. Experiments show that our proposed compression technique using the 2D interval wavelet transform can outperform JPEG2000 in terms of visual quality. Finally, a novel multi-layered wedgelet analysis is introduced to overcome the limitations and inefficiencies of wedgelet decomposition. Multi-layered wedgelet representation is able to reproduce geometrical image features at reasonable quality to facilitate visual interpretation at low bit-rate environments.