增广Lagrangian的快速视频复原方法

针对单幅模糊图像复原的局限性和视频应用的广泛 性,提出了一种基于时空体和增广Lagrangian的快速视频复原方法。首先对视频复原与图像 复原的特征进行比较和分析,研究视频复原的三维解卷积操作, 并对目前存在的视频复原方法的实现过程与性能进行分析和总结；然后在时空体的思想下, 通过时空联合 的各向同性全变分来控制时间误差和空间误差,并引入一种增广的Lagrangian方法完成全变 分规整化的难 题；最后通过求解Lagrangian形式的f子问题和u子问题实现全变分最小化难题,并对规整化 过程中的参 数进行研究与讨论,最终实现了视频的快速鲁棒复原。基于仿真图像和实际视频的实验结果 表明,本文方法 的性能在运行时间和视觉质量评价方面都要优于当前的其它方法,能够有效地实现图像和视 频的快速复原。

Augmented Lagrangian based fast video restoration approach

Aiming at the limitation of single blu r image restoration and the universality of video applications,a fast and robust video restoration approach based on space-time volume and augmented Lagrangian is proposed. Firstly,the characteristics of image restoration and video restoration are com pared and analyzed,the three-dimensional deconvolution of video restoration operator is studied,and the realized process and performance of the existing video restoration methods are analyzed and summarized.And then,on the basis of the idea of space-time volume,the temporal error and spatial error i s controlled through spatio-temporal isotropic total variation,and an augmented Lagrangian based method is introduce d to implement the total variation regularization problem.Finally,the problem of total variation minimization is achieved by solving the u sub-problem and f sub-problem in the form of Lagrangian,and the parameters are studied and discussed in the process of regularization,which achieves fast and robust video restoration at l ast.The experimental results based on both simulated images and real videos show that the performance of the propos ed approach is superior to the current state of the art methods in the aspect of running time and several visua l assessment indices,which can efficiently achieve the task of fast image and video restoration.