引入高阶P-Laplace的图像全变分盲复原方法-计算机工程与应用.PDF

220 2017 ，53（2 ） Computer Engineering and Applications 计算机工程与应用 引入高阶P-Laplace 的图像全变分盲复原方法 吴亚娟，徐黎明 WU Yajuan, XU Liming 西华师范大学 计算机学院，四川 南充 637009 College of Computer Science, China West Normal University, Nanchong, Sichuan 637009, China WU Yajuan, XU Liming. Total variation blind restoration using high-order P-Laplace. Computer Engineering and Applications, 2017, 53 （2 ）：220-224. Abstract ：To solve the problems of deterministic blind restoration, like the special phenomena of ringing and insufficient suppression noise, the high-order P-Laplace operator is introduced into the classical total variation restoration methods. Firstly, shock filter is used to predict the clear edges of images, and then deal with edge and internal smooth region of images in different forms respectively according to changing rules of the natural image pixels. Finally, it restores the images by split Bregman iteration and alternating minimization algorithm. The experimental results show that the improved algorithm proposed in the paper can alleviate the ringing phenomenon, protect edge information and ameliorate the effects of image restoration compared to a number of blind restoration algorithms in recent years. Key words ：deterministic blind restoration; high-order P-Laplace; total variation; split Bregman; alternating minimization 摘 要：针对确定性盲复原中的振铃、阶梯现象和噪声抑制不充分的问题，将高阶P-Laplace 算子引入到经典的全变分 盲复原方法中。首先利用shock 滤波器进行清晰边缘预测，后根据自然图像中像素的变化规律，对图像的边缘和平滑 区域进行不同形式的处理，最后采用分裂布雷格曼迭代和交替最小化算法对非约束扩散方程进行求解。实验结果证 明，与近年的一些盲复原算法相比，提出的改进算法能较好地缓解振铃现象，保护好边缘信息并改善图像复原效果。 关键词：确定性盲复原；高阶P-Laplace ；全变分；分裂布雷格曼；交替最小化 文献标志码：A 中图分类号：TP391.41 doi ：10.3778/j.issn. 1002-8331.1607-0125 1 引言 据PSF是否已知将图像复原分为图像非盲复原（Non-Blind 在图像获取、传输、压缩、解压、存储等过程中都易 Image Restoration ）和图像盲复原（Blind Image Resto- [1] ration ）。图像盲复原是在PSF 未知或已知信息较少的