Title：System for recovering details in image denoising
Country：China
Patent No.：201911307899.0
Legal Status：Under review
Inventor：Xiang Xie, Xuyang Wang, Guolin Li, Zhihua Wang
Assignee：International Graduate School at Shenzhen, Tsinghua University; Tsinghua University
Address：Tsinghua Campus, Xili University Town, Nanshan District, Shenzhen City 518055, Guangdong Province
Filing Date：2019-12-18
Issue Date：
Abstract：
The invention relates to a system for recovering details in image denoising, and the system comprises: an initial denoising module which is used for extracting preliminarily denoised image informationS from a noisy image y; a detail recovery module which is used for estimating a missing detail part D, and marking the estimated detail information as D<^>; an initial denoising module which is usedfor transmitting the preliminarily denoised image information S to the detail recovery module, and combining the detail-free image information S with the detail information to obtain a final denoisingresult, and the initial denoising module and the detail recovery module are realized through a deep neural network sub-network. According to the system, the detail definition of the denoised image can be remarkably improved, the identifiability of the image to human eyes is enhanced, the subjective quality of the image is improved, and meanwhile objective indexes including the denoised peak signal-to-noise ratio (PSNR) and the structural similarity (SSIM) are improved.
Patent Certificate：
PDF/Jpg