A Novel Modified Total Generalized Variation Technique for Image Denoising

Abstract

The process of image denoising is very crucial in the field of image processing, since noise tends to adversely affect both the appearance and structure of images that are applied in different fields of study such as science and engineering. In this research paper, we present a new hybrid image-denoising scheme called Modified Total Generalized Variation (MTGV) for the restoration of noisy images. The proposed methodology makes use of the BM3D model along with a Modified Total Generalized Variation model to achieve better noise reduction while retaining image features, edges and other details. In the proposed scheme, the BM3D model is first applied to reduce the noise in the observed image. The mean of the observed image and the one obtained after applying the BM3D model is computed, which is further used in the restoration process using the proposed MTGV model. In order to measure the effectiveness of the suggested technique, the computational experiments were performed under different noise levels. The proposed approach is quantitatively analyzed in comparison with the BM3D and TGV approaches based on the metrics of PSNR and SSIM as well as computation time. Moreover, visual comparisons and intensity profiles along the horizontal and vertical directions are provided for selected examples of both low-noisy and high-noisy images. From the results of the experiment one can see that using the new MTGV technique results in noise suppression, preserving structural and intensity characteristics of original images.