基于稀疏表示和块匹配的图像超分辨率重建方法  

Image Super-Resolution Reconstruction Based on Sparse Representation and Block Matching

作　　者：王芳[1] 赵少林[1] 朱长征[2] 

机构地区：[1]郑州升达经贸管理学院信息工程系,河南郑州451191 [2]中国矿业大学计算机科学与技术学院,江苏徐州221116

出　　处：《湘潭大学自然科学学报》2017年第1期82-86,共5页Natural Science Journal of Xiangtan University

基　　金：河南省重点科技攻关项目(152102210176)

摘　　要：针对图像的超分辨率重建问题,提出一种基于稀疏表示和块匹配的重建方法.首先,根据图像退化模型将HR图像退化成LR图像.然后,通过训练获得过完备字典,对图像重建问题进行稀疏表示,并对此进行求解.最后,利用重叠块技术缓解块效应,利用反向投影技术保证全局一致性,最终获得重建的HR图像.另外,将该方法扩展到多帧图像重建中,利用块匹配技术从多帧图像中选择出一个目标图像,以此实现重建.实验结果表明,该方法的重建图像具有较低的峰值信噪比.For the issue that the image super-resolution reconstruction,a reconstruction method based on sparse representation and block matching is proposed.Firstly,the HR image is degraded into LR image according to the image degradation model.Then,the image reconstruction problem is modeled by sparse representation with the complete dictionary,and this problem is solved.Finally,the block effect is alleviated by using overlapping blocks,and the reverse projection technique is used to ensure the global consistency,so as to obtain the reconstructed HR image.In addition,the method is extended to multi-frame image reconstruction,and a target image is selected from multi frame images by using block matching technique.Experimental results show that the proposed method has a low peak signal to noise ratio.

关 键 词：图像超分辨率重建 稀疏表示 块匹配 多帧图像 峰值信噪比 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]