•  
  •  
 

Abstract

Image denoising is an important area of computer vision. Rudin-Osher-Fatemi model based on a gradient is one of the simplest models used in image denoising to solve the problem of restoring the clear image. The challenge in solving this model is the non-differentiability of Total Variation function (TV-function) minimization. Image transmission is widespread over wireless systems, including the fifth generation (5G) cellular network. Transmission impairment can affect transmitted images, including noise, attenuation, and distortion. This study proposed a new smoothing technique to make the TV-function differentiable and smooth. The new smoothed function was used for de-noising images with the help of the gradient descent method in minimization. Two transmission systems were proposed, additive white Gaussian noise (AWGN) and 5G enhanced mobile broadband (eMBB), to evaluate the performance of the proposed approach. The denoising technique proved convincing over AWGN and 5G eMBB channel models, reducing the noise effect on the transmitted images. Compared with the noisy image, the gains achieved by the denoised image reached 7.4 and 4174.1 for peak signal-to-noise ratio and mean square error, respectively. These gains are achieved at the low and moderate signal-to-noise ratio regions, while at the high signal-to-noise region, the quality of the noisy and denoised images is almost the same.

Share

COinS