一种改进的总变分正则化图像盲复原方法

传统的图像复原算法多数采用最小均方误差作为图像复原效果的评价标准，很少考虑人的视觉感受。本文在总变分盲复原算法的基础上结合Weber定律和正则化方法，运用不同的迭代表达式：在模糊辨识阶段，采用总变分正则化算法进行辨识；在图像复原阶段，采用Weber定律和正则化方法相结合。正则化的选择充分考虑图像的细节保持和边缘增强。实验结果采用基于人眼视觉感受的图像评价标准来验证。实验证明该算法在未知点扩散函数的情况下不但能有效的抑制噪声和消除纹波现象，而且还能有较好的视觉效果。

A modified method for total variational regularization blind image restoration

The traditional algorithm of image restoration generally uses minimum mean squared error as the evaluation standard of image restoration erects） and it seldom considered poople） s visual experience. In this paper, we combine the Weber＇ s law and regularization method based on the total variational blind recovery algorithm.The key ideas is to employ different iterative expressions by adopting the total variational regularization algorithra to identify in fuzzy identification stage, and combining the Weber＇ s law and regularizafion method in the image recovery stage.The choice of regolarizafion fully considers the image detail preserving and edge enhancing. The experiment result is tested by the image evaluation criterion which base on human vision remains feeling. The experiment result shows that this algorithm can not only effectively restrain noises and eliminate the phenomenon of the ripple, but can also have a better visual effect under the condition of unknown point spread function.