基于组稀疏表示的在线单帧图像超分辨率算法  

Online Single Image Super-resolution Algorithm Based on Group Sparse Representation

作　　者：李键红 吴亚榕[2] 吕巨建 LI Jian-hong;WU Ya-rong;LV Ju-jian(Laboratory of Language Engineering and Computing,Guangdong University of Foreign Studies,Guangzhou 510006,China;Department of Science and Technology,Zhongkai University of Agriculture and Technology,Guangzhou 510225,China;School of Computer Science,Guangdong Polytechnic Normal University,Guangzhou 510665,China)

机构地区：[1]广东外语外贸大学语言工程与计算实验室,广州510006 [2]仲恺农业工程学院科学技术处,广州510225 [3]广东技术师范学院计算机学院,广州510665

出　　处：《计算机科学》2018年第4期312-318,共7页Computer Science

基　　金：广东省科技计划项目(2016A020210131;2016A070712020);语言工程与计算实验室项目(LEC2016ZBKT004)资助

摘　　要：基于稀疏表示的图像超分辨率重建算法以近似随机抽取的方式选取字典中的原子来拟合图像片,而实际中的字典原子的选择体现出了很强的结构稀疏性,从而导致算法计算复杂且引入了大量的误差,影响重建图像的质量。针对该问题,提出了一种基于组稀疏表示的在线图像超分辨率重建算法。该方法引入组稀疏理论,仅利用输入的低分辨率图像作为样本来构建组稀疏字典,通过结合组稀疏性和几何对偶性来构建超分辨率图像算法的成本函数,并使用提出的一种迭代的方法进行求解。实验表明,该算法在视觉观察和参数比较上都优于当前主流的超分辨率算法。The super-resolution algorithm based on sparse representation selects atoms in the dictionary by approximately random style to fit the specified image patch,but the selected atoms show strong structure sparsity in practice,which leads to the complexity of calculation and a great deal of errors,affecting the quality of the reconstructed images.For conquering the problem,a online single image super-resolution algorithm based on group sparse representation was proposed.The proposed algorithm takes advantage of the inputted low-resolution image to construct the group sparse dictionary by introducing the group sparse theory,and then incorporates the group sparse prior and geometric duality prior to design the cost function of the algorithm,which is solved by a proposed iterative optimization method.The experiments demonstrate that the proposed algorithm is superior to the main stream algorithms subjectively and objectively.

关 键 词：组稀疏表示 单帧图像超分辨率 正交匹配追踪 字典学习 迭代优化 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]