Motion image restoration based on sparse representation and guided filter

Abstract

When moving objects are present, current low-resolution blurring image reconstruction techniques with considerable noise do not perform well. This paper comes up with a new image reconstruction method based on K-SVD algorithm and guided filter technique. This method uses K-SVD to pre-process the image first and apply canny boundary detector to obtain clear boundaries as prior model, thus we can estimate blurring kernel. Last, we apply guided filter to reconstruct our image. We do the second and third step iteration to obtain clear images. This paper uses simulated degeneration and actual low-resolution blurring image for experiments and our result implies this method has good performance for reconstruction.