SPOS: Deblur image by using sparsity prior and outlier suppression

In this paper, we propose an effective and robust method SPOS (Sparsity Prior and Outlier Suppression) for blurry image restoration. First, we combine histogram equalization and prior constrain to obtain the salient structure which contains main texture of image. Second, kernel is estimated by salient structure and sparsity constrain. Final, the blurry image is restored by non-blind deconvolution. The contributions of SPOS lie in two aspects: (1) we combine histogram equalization and sparsity suppression to obtain salient structure; (2) we take kernel outliers into consideration and introduce L-0 norm to suppress kernel's shape. The experiment results show that SPOS has the better performance compared with the state-of-the-art methods.