基于L0范数和联合全变分的图像去模糊  

L0 norm and combined total variation based image deblurring

作　　者：卢鹏 余勤[1] LU Peng;YU Qin(College of Electrical Engineering,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学电气工程学院,四川成都610065

出　　处：《计算机工程与设计》2020年第8期2284-2288,共5页Computer Engineering and Design

摘　　要：为解决图像因各种干扰而模糊的问题,基于典型的正则化约束方法,提出两种图像去模糊正则化模型的图像重建方法。在图像非盲去模糊方面,提出一种基于L0范数和非局部全变分(NLTV)的图像非盲去模糊模型,对图像平滑区和非平滑区分别采用L0范数和NLTV约束,对两个部分采用不同的算法分别求解;在图像盲去模糊方面,提出一种基于L0范数和联合全变分的图像盲去模糊模型,利用L0范数约束求解和模糊核估计方法得到模糊核和潜在图像,用联合全变分对潜在图像求解。实验结果表明,两个模型均有较好的去模糊效果和一定的抗噪鲁棒性。To solve the problem of image blurring due to various interferences,based on the typical method of regularization constraint,two image reconstruction methods of image deblurring regularization model were proposed.In the aspect of image non-deblurring,an image non-blind deblurring model based on L0 norm and non-local total variation(NLTV)was proposed,the L0 norm and NLTV regularization were used to constrain the smooth and non-smooth areas of the image respectively,and different algorithms were applied to solve the two sub-problems respectively.In terms of image blind deblurring,an image blind deblurring model based on L0 norm and joint total variation was proposed,blurring kernel and latent image were obtained using L0 norm constraint and blurring kernel estimation method,and the joint total variation method was used to deal with the latent image.Experimental results show that both models have good deblurring effects and certain robustness to noise.

关 键 词：L0范数 联合全变分 图像去模糊 交替方向乘子法 正则化先验 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]