Image compression using M- adic wavelets

Abstract

In this paper, we propose a low bit rate image coding scheme using M-adic wavelet transform. M-adic wavelets arise from solutions to the dilation equation, ¢(x) = Lk ¢(Mx - k). The M-adic wavelet transform of an image is computed using tree structured perfect reconstruction filter banks. The transformed image is coded into a compressed bit stream using a human visual system (HVS) model and a wavelet image model. HVS helps in removing perceptual redundancies, while the wavelet image model provides a framework to exploit the redundancies across different scales of the wavelet-transformed image. Two wavelet image models, namely, zerotree model and web model are generalized to be applicable to M-adic wavelets. Using these models the paper describes algorithms for obtaining an embedded bit stream. Simulation results show that the proposed algorithms can cater to a wide range of applications.