Alpha稳态噪声下基于Meridian范数的全变分图像去噪算法  

A Total Variational Approach Based on Meridian Norm for Restoring Noisy Images with Alpha-stable Noise

作　　者：杨真真[1,2] 杨震[2] 李雷[1] 金正猛[1] 

机构地区：[1]南京邮电大学视觉认知计算与应用研究中心,南京210023 [2]南京邮电大学宽带无线通信与传感网技术教育部重点实验室,南京210003

出　　处：《电子与信息学报》2017年第5期1109-1115,共7页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61501251;61271335;61271240);江苏省自然科学基金项目(BK20140891);南京邮电大学引进人才科研启动基金资助项目(NY214191)~~

摘　　要：在实际应用中,噪声不可避免,因此,图像去噪一直是图像处理领域研究的重点,并且近年来受到越来越多的研究者的青睐。该文首先基于Meridian分布和全变分(Total Variational,TV)的统计特性,提出一种全变分模型来复原alpha稳态噪声环境下的含噪声图像。此外,为了保证模型解的唯一性,对提出的全变分模型添加了一个二次惩罚项,得到一个严格凸的全变分模型,然后,使用原始-对偶算法对提出的全变分模型进行求解,并证明了该算法的收敛性。最后,进行了仿真实验,并对实验结果进行了分析,实验结果验证了提出模型的可行性与有效性。In actual applications, noises may inevitably exist, and thus to study the denoising method for images is great significant task in image processing filed that attracts much attention in recent years. In this paper, based on the statistical property of Meridian distributed and the Total Variational （TV）, a total variational method is proposed for restoring images degraded by alpha-stable noise. Besides, in order to obtain a strictly convex model, a quadratic penalty term is added, which guarantees the uniqueness of the solution. For solving the novel convex variational model, a primal-dual algorithm is employed to solve the above model, and the convergence of the algorithm is proved. The experimental results demonstrate that the feasibility and effectiveness of the proposed model for the noisy images with alpha-stable noise.

关 键 词：图像处理 Meridian范数 Alpha稳态噪声 原始-对偶算法 全变分 

分 类 号：TN911.73[电子电信—通信与信息系统]