一种MRF-MAP框架下的图像超分辨率重建方法  

A Method of Images Super-resolution Reconstruction Based on MRF-MAP Frame

作　　者：鲁林[1] 李翠华[1] 张珍[1] 余礼钹[1] 张东晓[1] 施华[1] 

机构地区：[1]厦门大学信息科学与技术学院,福建厦门361005

出　　处：《厦门大学学报（自然科学版）》2012年第4期696-700,共5页Journal of Xiamen University：Natural Science

基　　金：国防基础科研计划项目;国防科技重点实验室基金;福建省自然科学基金项目(2011J01365);高等学校博士学科点专项科研基金项目(20110121110020)

摘　　要：基于多帧观察样本的超分辨率图像重建是超分辨率图像重建研究中的重要方向.在马尔科夫随机场-最大后验概率(MRF-MAP)框架下研究了多帧图像的超分辨率重建问题.根据给定的空间图像退化模型建立了超分辨率重建的二阶能量函数,并利用α-expansion图切算法对能量函数进行求解.考虑到α-expansion算法的规范性要求,将能量函数进行近似.针对二阶能量函数的图切算法,讨论了s-t图的构造,给出一种节点的分配方法以及t-link和n-link的赋值方式,以提高图切算法的计算效果.通过对两种类型的图像进行超分辨率重建的对比实验,表明该方法具有较好的去噪及重建效果.Super resoltion based on multi-observation play a impostant role in the studing of subar-resoluting. This paper focus on su-per-resolution on images based on the frame of MRF-MAP. We construct the two clique energy function according to the images ob servation model. After a brief introduction to graph-cut, we use a-expansion algorithm to calculate the optima of energy function. Con-sidering the regularity that a expansion algorithm have on energy function, we approximate the energy function. In order to improve the quality of the reconstruction result, we design the construction of s-t graph for graph-cut algorithm and the assignment of the weight of the t-link and n-link. Via comparative experiment using two different type of pictures for super-resolution reconstruction, the method in this paper do better on super-resolution and denoising.

关 键 词：超分辨率重建 马尔科夫随机场 最大后验概率 图切算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]