A Biological Image Restoration Method with Independently Local Dictionary Learning

Abstract

In Recent years, sparse representation has been significantly applied in many image processing problems, such as image restoration, face recognition and image super-resolution, and shown promising results. The key issue of sparse representation is how to find a reasonable representation dictionary, through which the image can be presented more sparsely. In this paper, we addressed the biological image restoration, which was an important preprocessing technique for security identification, and proposed a novel sparse-based cost function. Considering the significant difference of underlying structure within different patches, we independently trained the dictionary using a set of self-similarity patches to present each patch more sparsely. To solve the proposed cost function, an approach based on alternating optimization was presented to obtain the approximate solution. Some experiments on face and palm images demonstrated that the proposed method was superior to many existing excellent methods.