基于Tikhonov和全变分正则化混合约束盲去模糊方法

Tikhonov and total variation regularization method for blind image restoration

为了使模糊图像在边缘区域和平滑区域有更好的复原效果,通过分析图像的稀疏平滑特性,提出一种基于Tikhonov和全变分正则化混合约束（Tikhonov-Total variation model,TTV）模型的盲去模糊方法。首先用Shock-Direction Filter（简称为SDF滤波器）提取图像边缘信息,然后利用Tikhonov模型和全变分模型分别在图像特殊区域的不同表现,分别对不同区域进行不同的正则化约束,最后提出一种多变量正阈值约束的分裂布雷格曼（Multivariable Positive threshold constraint Splitting Bregman,MPSB）算法对提出的新型模型进行数值最优化求解。从实验结果可以看出,相对对比算法,该方法不仅能够准确的估算模糊核（Blurring Kernel,BK）,而且在主观上实现了更好的复原效果,在客观上本文方法的峰值信噪比（PSNR）增量提高了3.4~4.7 d B。

In order to make the fuzzy image have a better restoration effect in the edge region and the smooth region, by analyzing image sparse smooth characteristics of images a method is presented based on Tikhonov and total variation point mixed regularization constraint （Tikhonov-Total variation, TTV ） model of blind deblurring. Firstly, Shock-Direction filter （ SDF filter） is used to extract the edge information of the image. Then, the Tikhonov model and total variation model are used in the special area of images on different performances , different regularization constraints in different regions, respectively. Finally, a multi-variables threshold constraint of the split Bregman （multivariable positive threshold constraint split- ting Bregman, MPSB） algorithm is used for solving numerical optimization of the model. Experimental results show that compared with the existed algorithms, the method can accurately estimate the fuzzy ker- nel （Blurring kernel, BK） and achieve the better restoration effect in the subjective, the peak signal to noise ratio（PSNR） increases 3.4 -4.7 dB.