Overview of traditional denoising and deep learning-based denoising

Abstract

Image denoising is a classical problem in the current field of computer vision. The goal of the task of image denoising is to use techniques to preserve as much clear detail of the original image as possible when the image has external noise. The essence of the image denoising process is to reduce the noise in the digital image and to recover and reconstruct the original clear image. The reason for image noise is that during image transmission and acquisition, the integrity of the image cannot be guaranteed due to environmental, acquisition equipment, human and other factors, so the image will inevitably be damaged by different degrees of noise. In medical, military, and optoelectronics fields, there is an extremely high demand for image realism, so the task of image denoising becomes very important. In this paper, we will discuss a history of image denoising techniques and analyze the denoising methods into traditional denoising and deep learning based denoising.

Image denoising via multi-scale spatial frequency domain convolutional neural network Noise-Net

“…In order to solve the problems that may occur in traditional filtering methods, in recent years, the research focus in the field of image denoising has gradually shifted to the development of deep learning image denoising algorithms because of deep learning and its excellent performance in the field of image processing [14,15] . CNN-based image denoising methods mainly focus on the extraction of feature information from the convolutional layer and the optimization of the network structure.…”