基于参数估计的降晰函数辨识及图像复原算法

成像系统的点扩展函数(PSF)以及观测噪声,在一般应用过程中是未知信息,因此,点扩展函数的辨识是一个具有挑战性的世界难题.为解决实际工作中遇到的在已知降晰类型情况下的降晰函数辨识和降晰图像复原问题,提出了基于参数估计的降晰函数辨识及降晰图像复原算法.首先,由初始猜测给定降晰函数参数的变化范围和参数的增量步长;然后,最小化降晰图像和由相应点扩展函数及降晰图像得到的实验观测图像的差的Frobenius范数,以确定点扩展函数的参数,进而确定降晰图像的点扩展函数并对降晰图像进行复原.应用基于Wiener滤波的频域循环边界算法对降晰图像进行复原.实验结果表明:在降晰图像信噪比较高的情况下,降晰函数的辨识结果是可靠和准确的,有较好的复原效果.

Algorithm of blur identification and image restoration based on parameter estimation

The point spread function(PSF)of the imaging system and the observation noise.are unknowna priori information in general applications.The identification of the PSF is a challenging and difficult problem in the world.In order to solve the problem of the identification and restoration when the degradation type is known,the algorithm of identification of the PSF and the restoration of the blurred images based on parameter estimation is proposed.First,the changing scope and the increment step length of the parameters are provided based on the original estimation.Second,the criterion in which the Frobenius norm of the difference between the estimated image with the corresponding PSF and the blurred image is minimized in every iteration step,and incorporated in order to determine the parameter of the PSF.Therefore,the PSF can be identified with the estimated parameter and the original image can be estimated via the general image restoration algorithms.In this paper,the frequency domain restoration algorithm based on the Wiener filtering is applied to restore the original images.The experimental results show that the identified result of the PSF is reliable and accurate,and the restoration effect via the identified PSF is better when the degraded image has high SNR.