基于学习字典和稀疏约束的湍流退化图像盲复原  

Blind restoration of degraded turbulent images based on learning dictionary and sparse constraints

作　　者：鱼轮 李晖晖[2] Yu Lun;Li Huihui(School of Electronic Information and Electrical Engineering,Shangluo Un iversity,Shangluo 726000,China;L aboratory of Inform ation Fusion Technology and Ministry of Education,College of Automation,Northwestern Polytechnical University,Xi’an 710129,China)

机构地区：[1]商洛学院电子信息与电气工程学院,商洛726000 [2]西北工业大学自动化学院信息融合技术教育部重点实验室,西安710129

出　　处：《电子测量技术》2020年第17期105-109,共5页Electronic Measurement Technology

基　　金：国家自然科学基金资助项目(61333017)资助。

摘　　要：针对基于字典稀疏表示图像复原中,利用清晰图像库训练样本导致的字典学习效率低,以及解卷积方法对噪声敏感的问题,提出一种基于学习字典和稀疏约束的湍流退化图像盲复原算法,该算法对观测图像进行降采样作为训练样本,利用字典下模糊图像与清晰图像的稀疏表示系数相同,直接用清晰字典和稀疏系数对图像进行复原避免解卷积运算。在利用稀疏系数复原图像时,加入图像的梯度稀疏先验,有效减轻了复原图像中的块效应。仿真结果表明,算法能在一定程度上抑制复原图像中的伪像,提高图像的复原效果。Aiming at the problems of dictionary based sparse representation image restoration,the problem of low dictionary learning efficiency caused by the use of clear image library training samples and the sensitivity of the deconvolution method to noise,this paper proposes a turbulence degraded image blind restoration algorithm based on learning dictionary and sparse constraints.The algorithm downsamples the observed image as a training sample,uses the sparse representation coefficients of the blurred image and the clear image under the dictionary,and directly restores the image with the clear dictionary and sparse coefficients to avoid deconvolution operations.When restoring an image using sparse coefficients,adding the gradient sparse prior of the image effectively reduces the blockiness in the restored image.The simulation results show that the algorithm can suppress the artifacts in the restored image to a certain extent and improve the restoration effect of the image.

关 键 词：图像复原 学习字典 点扩散函数 稀疏约束 

分 类 号：TN911.73[电子电信—通信与信息系统]