基于广义递归反演的超分辨率图像恢复研究  

Research on Super Resolution Image Restoration Based on Generalized Recurrent Inference

作　　者：赵云龙[1] 徐帅[1,2] 李光[3] 

机构地区：[1]西南交通大学信息科学与技术学院,四川成都610031 [2]西南交通大学峨眉校区,四川峨眉614202 [3]北京北广科技股份有限公司,北京100016

出　　处：《通信技术》2008年第11期182-184,共3页Communications Technology

摘　　要：超分辨率图像恢复在安全监控、遥感监测、军事侦察等领域有着广泛的应用。文中提出了一种基于广义递归反演的超分辨率图像恢复算法,该算法引入正则化因子,将超分辨率图像恢复的欠定问题转化为适定问题,然后通过递归的方式来恢复超分辨率图像。仿真结果表明,该算法能从多帧低分辨率图像来恢复所需要的超分辨率图像,而且算法是稳健的和有效的。Super resolution image restoration is extensively applied in security monitoring, remote sensing monitoring, military reconnoitering and so on. Based on generalized recurrent inference, this paper proposes an algorithm of super resolution image restoration, which turns under-constrained problem, namely super resolution image restoration, into well-pose problem by introducing regular factor, and then restores super resolution image in recurrent way. Simulation results show that this algorithm can restore super resolution image from several low resolution images, and it is robust and effective.

关 键 词：图像恢复 超分辨率 广义递归反演 正则化 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]