基于二阶广义全变差正则项的模糊图像恢复算法  

Blurred Image Restoration Method Based on Second-order Total Generalized Variation Regularization

作　　者：任福全[1] 邱天爽[1] 

机构地区：[1]大连理工大学电子信息与电气工程学部,大连116024

出　　处：《自动化学报》2015年第6期1166-1172,共7页Acta Automatica Sinica

基　　金：国家自然科学基金(61172108;61139001;81241059);国家科技支撑计划基金(2012BAJ18B06)资助~~

摘　　要：针对图像去模糊问题,采用二阶广义全变差作为修复图像的正则项构建恢复模型,并针对重建模型的高阶与非光滑特性,给出了基于分裂Bregman迭代的快速算法.实验结果表明,该模型和数值算法能够较好地恢复被噪声和模糊污染的图像,同时可以很好地保留图像的纹理和细节信息.For the blurred image restoration problem, we adopt the second-order total generalized variation as the regularization term to construct an image restoration model. For the high order and non-smooth feature of the restoration model, a fast algorithm based on the split Bregman iterative algorithm is also proposed. Experimental results show that the model and the numerical algorithm can effectively restore the images polluted by noise and blur, and they can preserve image texture and details effectively.

关 键 词：二阶广义全变差 图像恢复 去模糊 分裂Bregman迭代 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]