基于多成分字典和稀疏表示的超分辨率重建算法  

Super-resolution reconstruction algorithm based on multi-component dictionary and sparse representation

作　　者：刘梓[1] 宋晓宁[1] 於东军[1] 唐振民[1] 

机构地区：[1]南京理工大学计算机科学与工程学院,江苏南京210094

出　　处：《南京理工大学学报》2014年第1期1-5,共5页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(90820306);江苏省自然科学基金(BK2011492);中国博士后科学基金(2011M500926);江苏省博士后科学基金(1102063C)

摘　　要：为了解决单幅图像超分辨重建的问题,该文提出了一种在稀疏表示理论框架下基于多成分字典的方法。首先根据图像的退化模型,深入分析高低分辨率图像的关系,得到高分辨图像可由低分辨率图像在对应字典下的稀疏系数来重构的结论。根据这一结论,采用多成分字典分别表示图像的不同结构特征,并用匹配追踪的方法得到低分辨图像在多成分字典下的表示系数,然后在对应的高分辨字典下对高分辨率的图像进行重建,实现了基于多成分字典的单幅图像超分辨率重建。该文所提的方法对单幅图像的超分辨率重建具有较好的通用性,相对于传统的超分辨重建算法,在自然图像和卡通图像的实验中验证了算法的有效性。In order to solve the problem of super-resolution reconstruction of single image,a hybrid approach is presented under the framework of sparse representation with multi-component dictionary. According to the image degradation model,the algorithms focus on the relationship between the low-resolution images and high-resolution images. This paper concludes that the high-resolution images can be reconstructed by the coefficients of low-resolution images in the corresponding dictionary. The sparse coefficients are obtained by the method of match pursuit based on the multi-component dictionary which indicates different structural characteristics of the image. The high-resolution images are reconstructed in the corresponding high-resolution dictionary. This paper introduces an objective and new strategy capable of efficiently guiding the image restoration. Extensive experimental studies conducted on the nature and cartoon images show the effectiveness of the proposed method.

关 键 词：超分辨率 稀疏表示 多成分字典 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]