中文摘要：

在数字图像去模糊的正则化求解过程中，为了更好地保持图像的边缘和纹理，抑制振铃效应，需要结合图像不同空间位置的信息，自适应地调节正则化参数。通过引入Abdou算子计算图像中每个像素的梯度幅值，并考虑人类的视觉系统特性，构造出空间域上的加权矩阵s，从而对正则化参数自适应加权，并采用共轭梯度法进行去模糊求解。与Lagendijk提出的基于局部方差构造加权矩阵的复杂计算不同，基于Abdou算子求解的方法简单可行，去模糊效果良好。

英文摘要：

During digital image deblurring,the local information of the image is incorporated into regularization to adaptively adjust the the regularization parameter for better preserve the edge and texture of the image and suppress ringing effects.This paper calculates the gradient amplitude of each pixels by introducing Abdou operator,constructing the weighting matrix s in the spatial domain with the consideration of human visual system.And the final step is to solve the problem by the method of conjugate gradient.Compared with the method proposed by Lagendijk which is based on local variance,the method based on Abdou operator is feasible and quite simple with good results.