基于 Tikhonov 正则的图像盲复原算法-《计算机技术与发展》

文章信息/Info

Author(s):: YANG Tian-ji; JIANG Ya-qin; GUO Xiao-ya; School of Science,Nanjing University of Posts and Telecommunications,Nanjing 210023,China

Keywords:: variational method; blind restoration; pyramid strategy; Tikhonov regularization; alternating minimization

摘要:: 图像在获取、传输及保存的过程中,很多因素会导致图像质量退化,图像模糊是图像质量退化的一种常见表现。基于全变差( TV) 的图像复原 Chan 模型虽然能较好地刻画导致图像质量退化的模糊核,但该模型的图像复原结果严重依赖于参数的选取。针对 Chan 模型对参数敏感的问题,在该模型中引入模糊核的 Tikhonov 正则,提出新的盲去糊模型,并证明新的盲复原模型解的存在性。另外,采用由粗到精的多层图像金字塔策略,构造模糊核的初始值,再结合交替极小化(alternating minimization,AM) 方法,设计基于初始模糊核的快速算法求解所提模型。数值实验结果表明:所提模型与其他正则化模型相比,在不需要模糊核动态阈值约束的前提下,不仅能得到高质量的图像复原结果,而且对参数有较好的鲁棒性。

Abstract:: Image degradation is influenced by many factors during image acquisition, transmission and preservation. Image blur is acommon representation of image degradation. Blur kernel leads to the degradation of image quality. Although the Chan model based ontotal variation ( TV) can estimate the blur kernel,the restoration result of this model depends heavily on the selection of parameters. In this paper,a new blind image deblurring model is proposed by introducing the Tikhonov regularization into Chan model in order to overcome the effect of parameter sensitivity. Then,the existence of solution of the new blind restoration model is proved. Besides,the initial value of the blur kernel is constructed using a multi-layer image pyramid strategy from coarse to fine and a fast algorithm based on the initial blur kernel is designed combining alternating minimization? ? ? ?( AM) method for solving the proposed model numerically. On the premise of without the constraint of blur kernel dynamic threshold,numerical results show that the proposed model can not only get high quality image restoration results,but also has stronger robustness for parameters compared with other regularization models.