加权变分的图像去噪算法.pdfVIP

32 2 V ol. 32 No . 2 2010 2 Sy stems Engineer ing and Electr onics Febr uar y 2010 : 1001506X( 20 10 1, 2 1 1 1 陈利霞 , 冯象初 , 王卫卫, 宋国乡 ( 1. 西安电子科技大学理学院, 陕西 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 : 针对经典的总变分去噪模型边缘信息对噪声敏感且 模糊的缺陷, 提出了非线性与线性的加权变分 模型非线性加权变分模型是在总变分模型的正则项中引入权函数, 并利用权函数引导扩散, 使得新模型在消噪 的同时更好地保持图像的纹理特征和边缘信息; 线性加权变分模型是对含噪图利用高斯函数进行预处理, 再对处 理后的图像进行扩散, 从而降低计算复杂度数值实验表明, 与经典的总变分模型相比, 改进的方法无论是在视 觉效果还是峰值信噪比上都有明显的提高 : 偏微分方程; 总变分; 图像去噪; 权函数 : T P 391 : A Image denoisi ng algorithms based on weighted vari ation 1, 2 1 1 1 CH EN L ix ia , FEN G Xiangchu , WA N G Weiw ei , SON G Guox iang ( 1. School of Science, X idian Univ., X ian710071, China; 2. School of Mathematics and Comp uting Science, Guilin Univ. of lectronic Technology , Guilin541004, China) Abstract: View on the w eakness of a classical variational denoising m odel in w h ich edge inform at ion is sen sitiv e to noises and prone to blur, tw o impr ov ed nonlinear and linear w eight ed v ar iational algorithm s ar e put for w ard. In the nonlinear w eighted v ar iat ional model, a w eight function is intr oduced in the regular izat ion t er m of t he classical m odel t o induct diffusion, w hich giv es the result that the new model pr eserves the textur e chara ct er istics and the edge infor m ation bet ter w hile removing noises. In t he linear model, the Gaussian function is used t o sm ooth the noised image be