Reliable Image Enhancement Based on Evidence and Generative Adversarial Networks
Zhonghai He, Xiang Dong Xu, Zhuming Lian, Yu Luo · 2024
Deep learning have been widely applied in image enhancement, but traditional deep learning methods have not taken into account the uncertainty and reliability of image enhancement. We propose a new method of using generative adversarial networks to enhance image details, and introduce evidence theory to consider eliminating uncertainty during the training process of Generative Adversarial Networks. The simulation experiment on COVID-19 image data shows that the proposed method can supplement the details between high- quality images and low-quality images, improve the classification accuracy after image enhancement, and effectively increase the reliability of image enhancement.