Image Denoising Based on A New Wavelet Statistical Model
Xili Wang, Xiyuan Wang, Han Cao · 2006
A new image denoising algorithm is proposed. It models wavelet coefficients as Laplacian distribution. An exponential type a prior distribution is defined for the Laplacian distribution parameter. The image wavelet coefficients can be regarded as a realization of a doubly stochastic process. Then one can use maximum a posterior (MAP) estimator and spatial adaptive technique to estimate the distribution parameter and the clean coefficients. The method is realized with an over complete transform known as dual tree complex wavelet transform (DT-CWT). Experiments demonstrate the effectiveness and low complexity of the new algorithm