基于L_1范数的总变分正则化超分辨率图像重建  

L_1 Norm of Total Variation Regularization Based Super Resolution Reconstruction for Images

作　　者：占美全[1] 邓志良[2] 

机构地区：[1]江苏科技大学计算机科学与工程学院,镇江212003 [2]常州信息职业技术学院,常州213164

出　　处：《科学技术与工程》2010年第28期6903-6906,共4页Science Technology and Engineering

摘　　要：设计了一种基于L1范数的总变分正则化超分辨率图像序列重建算法。采用L1范数对重建图像保真度进行约束,利用总变分正则化克服重建问题的病态性,有效地保持了图像的边缘并且提高了运算速度;运用设计的算法对模拟的低分辨率图像序列进行重建,分别从主观效果和客观衡量指标两方面与基于L2范数的总变分正则化的超分辨率重建结果进行比较,实验结果表明该算法在保持图像边缘的同时,提高了超分辨率重建算法的运算速度。An L1 norm of total variation regularization based super resolution reconstruction algorithm for images was proposed.The L1 norm was used to constrain the fidelity of the reconstructed image,and the total variation regularization was implemented to overcome the ill-posed of the problem.The edge of the image was preserved effectively and the speed of the algorithm was improved.Simulated taken low resolution image sequences for the designed algorithm were used in the experiments,and compared the proposed algorithm with the algorithm that based on L2 norm of total variation regularization for super resolution reconstruction in the way of subjective vision effect and objective quality,respectively.The results show that the proposed algorithm not only preserves the edge of image,but also improves the computing speed of the algorithm for super resolution reconstruction for image sequences.

关 键 词：总变分 正则化 超分辨率 L1范数 L2范数 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]