WGAN-Based Image Denoising Algorithm

1 Related Work

The combination of deep learning has a significant impact on image denoising. The main methods include multilayer perceptron (Rosenblatt, 1959), automatic encoder (Rumelhart, Hinton, & Williams, 1985), and deep convolutional neural network (LeCun et al., 1989). Burger et al. (Harold Christopher Burger, Christian J Schuler, & Stefan Harmeling, 2012a, 2012b) used multilayer perceptron for image denoising, which had a better denoising impact than BM3D. In 2017, Zhang et al. (K. Zhang et al., 2017) proposed a deep convolutional neural Gaussian noise denoising network (DnCNN) based on residual learning. Its denoising result is better than that of other traditional denoising algorithms, and it is currently considered as one of the best denoising algorithms. Cheng Chung Ming at al. (C.-M. Chen, Zhang, & Hsu, 2020) developed a new model to achieve extraction of weak edges in speckle noisy images, the level set model called Edge Attraction Force. Kuo-Kun (Tseng et al., 2019) used Kalman filtering to estimate the state of a dynamic system based on a series of measurements that did not fully account for the noise.