Image Restoration In Sparse Domain Using Adaptive Dictionary With Regularization
Sparse representation is one of the powerful statistical image modeling techniques. Of late, sparse modeling
techniques have been widely used in the area of image restoration. The basic concept behind the idea is that natural images
are sparse in some of the domain. The quality of the recovered image using sparse modeling techniques greatly depends on
the sparsity of the domain. In this paper we propose an algorithm which adaptively chooses a sub-dictionary to represent
each patch within an image, such that the entire image will have a better sparsified representation. In addition, we introduce
an adaptive regularization term called smoothing regularization parameter. The smoothing regularization parameter adds an
extra constraint into the l2-minimization problem. The proposed algorithm gives much improved results than most of the
state of the art algorithms in terms of PSNR and sharpness.
Keywords—Image restoration,sparse modeling, Image deblurring, Regularization.