1 基于TV的小波修复模型 - 中国图象图形学报.DOC

中图法分类号:TP391.41 文献标识码:A 文章编号:1006-8961(2015) - 论文引用格式：Xu J L,Hao Y, Zhang J. Second order total generalized variational model for wavelet inpainting [J].Journal of Image Graphics, 2015, . [许建楼, 郝岩, 张冀. 二阶总广义变分小波修复模型[J].中国图象图形学报, 2015, .] [DOI:10.11834/jig.2015 ] 二阶总广义变分小波修复模型 许建楼， 郝岩， 张冀 河南科技大学数学与统计学院, 洛阳 471023 摘 要：目的：针对全变分小波修复模型易导致阶梯效应的缺陷, 提出了一种加权的二阶总广义变分小波修复模型。方法：不同于全变分小波修复模型，假设的新模型引入二阶导数项且能够自动的调解一阶和二阶导数项。另外,为有效的利用图像的局部结构信息,新模型引入了权函数，它即能保护图像的边缘的同时又增强光滑区域的去噪能力。 为有效的计算新模型,首先利用交替方向法将该模型变为两个子模型, 然后对两个子模型分别给出相应的理论和算法推导。结果：实验结果表明: 相比最近基于全变分正则小波修复模型(平均信噪比，平均绝对误差及平均结构相似性指标分别为21.8844,6.8578,0.8272),新模型得到更好的修复效果(平均信噪比，平均绝对误差及平均结构相似性指标分别为22.3138,6.6261,0.8318)。结论：与全变分正则相比，二阶总广义变分正则更好地减轻阶梯效应。目前, 国内外学者对该问题的研究取得一些结果。由于原始-对偶算法需要较小的参数，所以运算的速度较慢，因此更快速的算法理论有待进一步研究。另外，该正则能应用的图像去噪、分割、放大等方面。 关键词： 全变分；总广义变分； 图像修复；交替方向法；原始-对偶算法。 Second order total generalized variational model for wavelet inpainting Xu Jianlou，H Yan，Zhang Ji School of Mathematics and Statistics,Henan University of Science and Technology, Luoyang China, 471023 Abstract: Objective: Aimming at the drawback of staircase effect based on the total variational method, we proposed a weighted second order total generalized variational model for wavelet inpainting. Method: Unlike the method based on the total variation, the new model contains the second-order derivative term and the first-order derivative term, and it can automatically balance the first and the second order derivative terms through two regularization parameters. In addition, in order to make use of the local structure information of image, an edge indicator function is introduced in the proposed model. The edge indicator function tends to 0 when the pixels belongs to the edge domain of image, and when the pixels belongs to the smooth domain of image, the edge indicator function tends to 1. So the proposed edge indicator function can preserve the edges and small fine parts of restored image, meanwhile, it can improve the ability of noise rem