摘要
受强噪声污染的图像经过现有降噪算法降噪后,往往很难恢复出图像的细节,导致图像分辨率不高,严重影响图像后期的应用,所以研究受强噪声污染图像的高清细节恢复算法具有重要意义。提出了一种基于双段深度残差卷积网的强噪声超分辨率重建算法,该算法设计了一种双段深度残差卷积网,前后段深度残差卷积网的结构完全相同,前段深度残差卷积网用于降噪,后段深度残差卷积网用于超分辨率重建。结果表明,提出的算法可以很好地适用于强噪声的图像降噪,不仅能获得更高的峰值信噪比与结构相似度,而且能获得更好的图像细节恢复、改善图像的视觉效果,具有较好的实用性。
After the existing de-noising algorithm is used to de-noise the noise of images polluted by strong noise,it is often difficult to recover the image details,resulting in low image resolution,which seriously affects the later application of images.Therefore, it is of great significance to study the high-definition detail recovery algorithm of images polluted by strong noise.For this reason,this paper proposes a high noise super-resolution reconstruction algorithm based on double-segment deep residual convolution network.This algorithm designs a kind of double-segment deep residual convolution network.The structures of the front and the back segments of the deep residual convolution network are exactly the same.The deep residual convolution network is used for noise reduction in the front segment, and the deep residual convolution network is used for super-resolution reconstruction in the rear segment.Experimental results show that the algorithm proposed in this paper can be applied to image de-noising with strong noise.It can not only achieve higher peak signal to noise ratio(PSNR)and structural similarity(SSIM),but also obtain better image detail recovery and improve the visual effect of the image,which has good practicability.
作者
刘哲
刘政
王恩
LIU Zhe;LIU Zheng;WANG En(Institute of Civil-military Integration,Northwest University of Political Science and Law,Xi’an 710122,China)
出处
《武汉大学学报（工学版）》
CAS
CSCD
北大核心
2022年第3期300-309,共10页
Engineering Journal of Wuhan University
基金
国家自然科学基金项目(编号:61473237)
陕西省重点科研计划项目(编号:2017ZDXM-NY-088)。
关键词
深度学习
残差学习
图像降噪
超分辨率
deep learning
residual learning
image de-noising
super-resolution
