改进的总变分去噪算法  

Improved Total Variation Algorithms to Remove Noise

作　　者：陈利霞[1,2] 宋国乡[1] 丁宣浩[2] 王旭东[1] 

机构地区：[1]西安电子科技大学理学院,西安710071 [2]桂林电子科技大学数学与计算科学学院,广西桂林541004

出　　处：《光子学报》2009年第4期1001-1004,共4页Acta Photonica Sinica

基　　金：国家自然科学基金(10861005、10501009);广西自然科学基金(0542046)资助

摘　　要：提出两个改进的模型用以解决标准TV模型在处理纹理丰富的图像时易丢失重要信息以B用梯度检测边缘时易受噪音干扰的缺点.模型A是在标准TV模型的基础上作了两点改进,一是引入边缘检测函数来引导扩散,二是利用小波变换的模来检测边缘.这使得新模型不但根据图像的特征进行平滑,并具有较强的抗噪能力,从而能更好地保护边缘.模型B是基于噪音在小波域中的特性对模型A在计算复杂度上的简化.数值试验表明,这两个模型均比TV模型具有较好的性能.The standard TV model is prone to lose important information while processing images with rich textures, and generally, it uses the gradient, which is sensitive to noise, to detect edges. To overcome these weaknesses, two improved models are proposed. Based on the standard TV model, model A has two improvements： one is to introduce an edge detecting function to induct diffusion, the other is to use the module of a wavelet transform to detect edges. These improvements make model A not only can smooth according to the feature of the image but also can effectively suppress noise,therefore,it shows much better performance on preserving edges. Model B, which is proposed based on the feature of noise in wavelet domain,is a simplified version of model A on computational complexity. Numerical experiments show that both the proposed models outperform the standard total variation model.

关 键 词：偏微分方程 图像去噪 小波 总变分 TV模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]