Frame-based Image Denoising Using Local Contextual Hidden Markov Model

Abstract

Wavelet-domain hidden Markov models (HMMs) provide a powerful new approach for statistical modeling and processing of wavalet coefficients. However, the local statistics of the coefficients cannot be well characterized by the HMMs. In this paper, we consider the combination of LGMM and HMMs in frame-domain and propose a Local Contextual Hidden Markov model based on frame-domain and corresponding training algorithm. This model can exploit the local statistics of coefficients so that can effectively capture correlation of wavelet frame coefficients. To illustrate the power of this new model, we apply this model for image denoising that outperforms current image denoising methods.