A denoising inspired deblurring framework for regularized image restoration

Abstract

In this paper, we suggest a restoration scheme to approximate the true image degraded by motion or out-of-focus blur together with additive Gaussian noise. The upper bound on the use of regularization inspires image denoising prior to image deblurring. Further, noise removal depends on the precise knowledge of neighborhood statistics. Accordingly, an appropriate neighborhood around each test pixel is selected based on the noise variance and uncorrelated property of the additive noise. The lower bound of regularization is incorporated as an edge recovery constraint in the deblurring cost function. The suggested framework along with few existing schemes have been simulated on various standard images. The underlying PSNR metric validate the noise removal and edge preservation potential of our method over its counterparts.