题目: Image Denoising via Generative Adversarial Networks with Detail Loss
摘要: Image denoising is a challenging task which aims to remove additional noise while preserving all useful information. Many existing image denoising algorithms focus on improving the typical object measure, peak signal-to-noise ratio (PSNR), and take the mean square error (MSE) as their loss function. Although these algorithms can effectively improve the PSNR on the benchmark dataset, their denoised images often lose some important details or become over-smooth in some texture-rich regions. In order to solve this problem, we introduce the perceptual loss from single image super-resolution (SISR) filed into our image denoising work. Besides, for suppressing the generation of high-frequency artifacts, we additionally propose a new loss term which is named as high-frequency loss in this paper. To understand easily, we represent the whole loss, including the perceptual loss and the high-frequency loss, as detail loss. Replacing the MSE, the detail loss is taken as our loss function. Our experimental results show that our method outperforms the current state-of-the-art methods on preserving the details during denoising. Compared to those state-of-the-art methods, the denoised images by our method are clearer, sharper and more realistic on details.
专家简介：刘学峰，国家千人、科学院百人、教授，曾任最大光通信器件公司Bookham核心项目负责人，精密生物成像光学系统光学部门主管，最大半导体装备公司ASML最新光刻机型研发项目管理。现任南理工纳米显微光学技术研究所所长，主持国家“先进光电探测”学科创新引智基地（111计划)，主持国家自然科学基金面上项目，重大科研仪器专项，江苏省“双创计划团队”等多个项目，拥有相关核心专利 17 项。