正则图像恢复中正则化算子选取的定性分析

A Qualitative Analysis of the Selection of Regularization Operator in Image Restoration

对病态图像恢复中正则化算子的选取问题进行定性分析。以最小化正则解模糊误差为目标,利用泰勒级数定性地分析怎样选取正则化算子,得出结论:在信号的强度大于噪声强度的频带正则化算子应该是带阻的,而信号的强度小于噪声强度的频带正则化算子应该是带通的,通常情况下应选取低阻高通的正则化算子。实验结果表明了该结论的正确性。

The selection of regularization operator in image restoration is qualitatively analyzed in this paper.The focus is the minimization of the blurred error of regularized image.Taylor Expansion helps to choose regularization operator qualitatively,and yields a conclusion that in the frequency ranges of spectrum where the strength of image is greater than that of noise the regularization operator should be band-stop and in the frequency ranges of spectrum where the strength of image is smaller than that of noise the regularization operator should be band -pass and in the usual situation the regularization operator should be low-stop and high-pass.Numerical results illustrate that the conclusion from the qualitative analysis is right.