Fusion of Wiener Filtering and BM3D Denoising for Improved Image Restoration

The objective of image restoration work, or image processes, is to return an observed image Y, that has been corrupted with noise, to its original form. In other words, given an image that consists of noise and blurred content, we aim to find the original image. The non-blind image restoration, in particular, focuses on the recovery of a case of unknown images with application of an assumed known blur. Wiener filter is a very popular image restoration tool. It can be thought of as the optimal sift in the rooted space of the blurred image to produce the least number of artifacts due to wider blur. However, one disadvantage is the need-to-know anticorrelations of the blur, the anti-blurred image and the noise. This paper contains the implementation of such non-blind image restoration where Wiener parametric filtering is used with BM3D. In this stage, the parametric Wiener filter is first used to deconvolve the image in the frequency domain, and then the BM3D technique is employed. The performance of the developed algorithms gives quite interesting and quite optimistic results.