Blind noisy deblurring via hyper laplacian prior and spectral properties of convolution kernel

Blind deblurring attempts to recover the latent sharp image from a blurred one. Such task is a well-known ill-posed inverse problem and is therefore usually solved as a posteriori probability estimation, incorporating prior information on natural images. In this paper, we propose a general blind noisy deblurring model based on hyper Laplacian (HL) in gradient domain and kernel spectra prior. This model includes the non-convex HL prior term, so we first separate variables and then utilize general soft threshold (GST) and closed-form threshold formulas (CFTF) to solve the proposed model, respectively. Simulation results verify the efficiency and feasibility of the proposed method. The proposed model can be used to solve other problems, such as machine learning and sparse coding.