Regularization has been widely used in image restoration. However, selection of the regularization parameter and the regularization operator is not solved completely and is still the main difficulty for adaptively regularized image restoration. This paper presents a new approach to select the local regularization parameter and the local regularization operator. The local regularization parameter is selected according to the distribution of local noise value in degraded images, and the local regularization operator is selected according to anisotropic properties. Experimental results show that the proposed methods work well in the presence of many types noise.