引入统计先验的人脸图像恢复

Introducing Statistical Prior Knowledge to Face Image Restoration

将人脸形状和纹理的统计先验作为约束 ,引入经典正则化图像恢复算法框架 ,并给出迭代求解算法 ;同时 ,定义了反映图像模糊程度的边缘活动度 ,并在迭代的每一步中计算图像的边缘活动度 ,以确定在迭代求解过程中人脸先验对解进行约束的程度 由于人脸统计先验的约束以及引入边缘活动度来指导迭代求解过程 ,避免了由经典恢复算法得到的结果中会出现的振铃波纹 对实验结果的分析和主观感受表明

The statistical prior knowledge for the shape and texture of a face is formulated as a constraint energy term, which is incorporated into the regularized framework for image restoration. An iterative algorithm is given to provide a numerical solution. Besides, an edge active measure (EAM)is defined to describe the blurring nature of an image, which is evaluated at each step of the iteration process to determine the weight of statistical prior constraint. The unfavorable ringing effects are avoided due to the incorporated statistical prior and guidance of the evaluated EAMs. Subjective and objective comparisons of restoration results verified the effectiveness of proposed approach for image preservation and noise suppression.