Blind Image Deblurring via Salient Structure Detection and Sparse Representation

Abstract

Blind image deblurring algorithms have been improving steadily in the past years. However, most state-of-the-art algorithms still cannot perform perfectly in challenging cases, e.g., when the blurred image contains complex tiny structures or the blur kernel is large. This paper presents a new algorithm that combines salient image structure detection and sparse representation for blind image deblurring. Salient structures provide reliable edge information from the blurred image, while sparse representation provides data-authentic priors for both the blur kernel and the latent image. When estimating the kernel, the salient structures are extracted from an interim latent image solved by combining the predicted structure and spatial and sparsity priors, which help preserve more sharp edges than previous deconvolution methods do. We also aim at removing noise and preserving continuity in the kernel, thus obtaining a high-quality blur kernel. Then a sparse representation based \(\ell _1\)-norm deconvolution model is proposed for suppressing noise robustly and solving for a high-quality latent image. Extensive experiments testify to the effectiveness of our method on various kinds of challenging examples.