Sparse representation of a blur kernel for out-of-focus blind image restoration

摘要: Blind image restoration is a non-convex problem which involves of images from an unknown blur kernel. The factors affecting the performance this are how much prior information about and kernel provided what algorithm used to perform task. Prior on often employed restore sharpness edges image. However, no consensus present regarding use in restoring due complex blurring processes. In paper, we propose modelling as sparse linear combinations basic 2-D patterns. Our approach has competitive edge over existing methods because our method flexibility customize dictionary design, makes it well-adaptive variety applications. As demonstration, construct formed by patterns derived Kronecker product Gaussian sequences. We also compare results with those other state-of-the-art methods, terms improvement SNR (ISNR).