Blind Image Deblurring via a Novel Sparse Channel Prior

Blind image deblurring (BID) is a long-standing challenging problem in low-level processing. To achieve visually pleasing results, it of utmost importance to select good priors. In this work, we develop the ratio dark channel prior (DCP) bright (BCP) as an for solving BID problem. Specifically, above two priors obtained from RGB images are used construct innovative sparse at first, and then learned incorporated into tasks. The proposed enhances sparsity DCP. At same time, also shows inverse relationship between DCP BCP. We employ auxiliary variable technique integrate information iterative restoration procedure. Extensive experiments on real synthetic blurry sets show that algorithm efficient competitive compared with state-of-the-art methods blind effective.