基于L_(1/2)正则化的超分辨率图像重建算法  

Super-resolution Image Reconstruction Algorithm Based on L_(1/2) Regularization

作　　者：王欢[1] 王永革[1] 

机构地区：[1]北京航空航天大学数学与系统科学学院,北京100191

出　　处：《计算机工程》2012年第20期191-194,共4页Computer Engineering

基　　金：国家自然科学基金资助项目(10801007);国家"973"计划基金资助项目(2010CB731900)

摘　　要：为提高图像重建质量,研究超分辨率图像重建技术与稀疏表示理论,提出一种基于L1/2正则化的超分辨率图像重建算法。将L1/2正则化理论运用到字典学习中,利用学习得到的字典重建高分辨率图像。实验结果表明,该算法的图像重建效果优于基于L1正则化的超分辨率图像重建算法。In order to improve the image reconstruction quality,by studying the super-resolution image reconstruction technology and the theory of sparse representation,this paper proposes a super-resolution image reconstruction algorithm based on L1/2 regularization.It applies L1/2 regularization into dictionary learning,and reconstructs super-resolution images using learned dictionaries.Experimental results show that the reconstruction results in this paper are better than the results of super-resolution image reconstruction algorithm based on L1 regularization.

关 键 词：L1 2正则化 稀疏表示 超分辨率图像重建 K-SVD算法 字典学习 训练样本 

分 类 号：TN911.72[电子电信—通信与信息系统]