Efficient Method for High-Quality Removal of Nonuniform Blur in the Wavelet Domain

摘要: This paper presents a novel nonuniform deblurring approach, which defines the blur model and calculates regularized deconvolution in wavelet domain to achieve high efficiency accuracy simultaneously. Targeting computation efficiency, we derive wavelet-domain hierarchical model, can be calculated efficiently by exploiting sparsity property of natural images domain. Correspondingly, is incorporated into multilayer framework at each layer spatially varying step sizes are introduced further accelerate convergence algorithm. In addition advantages, proposed approach deals with intensely due intrinsic tight supportness basis. We conduct series experiments comparisons validate effectiveness our