一种基于正则化参数自适应选择的快速近似求逆的图像恢复新算法

A NEW FAST APPROXIMATELY INVERSE MATRIX IMAGE RESTORATION BASED ON ADAPTIVE CORRECTION OF REGULARIZATION PARAMETER

本文根据正则化恢复中正则化参数应具有的性质,提出了一种基于正则化参数自适应选择(ACRP)方案的新的空域迭代恢复算法。通过这种正则化参数自适应选择,无论迭代初值如何选择,正则化参数都可以自动修正到最优值。根据对称正定矩阵的特点,我们对它近似求逆矩阵,由这种近似逆矩阵而得出空域中快速近似求逆(PAIM)图像恢复算法。实验证明,这种新的恢复算法可以显著地缩短恢复时间,并有效地抑制恢复的振铃效应。

In this paper, based on the character of regularization parameter in regularized image restoration, the adaptive correction of regularization parameter (ACRP) is proposed to our new iterative restoration algorithm in space domain.By the ACRP,regularized parameter can automatic correct to appropriate value whatever the initial condition is. According to the character of symmetrical and positive definite matrix, we calculate its inverse matrix approximately and then the regularized fast approximately inverse matrix image restoration in space domain is elicited. The new image restoration algorithm can distinctly shorten the time of restoration and availably induce the ringing artifacts around sharp intensity transitions.