图像去模糊问题是一个典型的反问题。受制于反问题的内在约束，在其庞 大的解空间中寻找真解或者符合视觉习惯的解都非常困难。再加上观测过程中 引入的噪声，更是制约解的质量。 针对图像去模糊的反问题特性，从模型上来看，已有的方法主要集中在两 个大的方向：1）寻求更恰当的图像先验知识来构造更精确的先验模型。这些 先验经历了从光滑性、分片光滑性、梯度稀疏性等诸多特性的演变，在图像的 盲目和非盲目复原方法中都广泛应用。2）根据对观测噪声的分析设计更合理 的保真项。(The image deblurring problem is a typical inverse problem. Subject to the inherent constraints of the inverse problem, it is very difficult to find the true solution or the visual habit in its large solution space. Moreover, the noise introduced in the observation process restricts the quality of the solution. In view of the inverse problem of image deblurring, from the model point of view, the existing methods focus mainly on two major directions: 1) seek a more appropriate image prior knowledge to construct a more precise prior model. These priors have evolved from many properties, such as smoothness, piecewise smoothness, gradient sparsity, etc., and have been widely used in blind and non blind restoration methods. 2) based on the analysis of the observed noise, a more reasonable fidelity term is designed.)