基于最小二乘影像匹配的超分辨率重建  

Super-resolution Reconstruction Based on Least Squares Image Matching

作　　者：朱伟[1] 孙久运 刘凯凯 

机构地区：[1]中国矿业大学环境与测绘学院,江苏徐州221116 [2]江苏省第二地质工程勘察院,江苏徐州221000

出　　处：《地理与地理信息科学》2014年第2期116-119,共4页Geography and Geo-Information Science

基　　金：国家自然科学基金项目"数字摄影测量影像目标精密定位与精度估计研究"(41171343)

摘　　要：通过序列影像间的互补信息改善影像失真和退化的超分辨率重建,其关键是精确获取序列影像间的运动信息。该文讨论了一种基于最小二乘影像匹配的高精度运动估计方法,根据最小二乘影像匹配的高精度同名像点,获取同名点在序列影像间的运动信息,从而进行低分辨率序列影像的子像素级运动信息精确估计,据此进行超分辨率重建。利用一组模拟低分辨率序列影像进行的超分辨率重建验证结果表明:基于最小二乘法运动估计精度较高,采用迭代反投影法重建影像具有较好的视觉效果,该方法尤其适用于存在平移运动的影像序列的超分辨率重建。Super-resolution reconstruction is applied for preventing images from distortion and degradation by supplementing information in sequences images,the key step of this algorithm is obtaining accurate motion information in sequence images.In this paper,algorithm of high precision motion estimation based on least squares image matching is proposed.This algorithm bases on the precision homonymy point matching of least squares image matching,obtains the motion information of homonymy points in the sequence images for accurate estimating motion information of sub-pixel of low-resolution rate sequence images.Then high quality super-resolution reconstruction results could be obtained according to the motion estimation.The experiment,using a set of analog low-resolution sequence image,estimates the motion of image by the least squares algorithm,and then reconstructs by iterative back projection algorithm.The experiment result shows that reconstructed images have better visual effects for high precision of motion information,particularly for super-resolution reconstruction of image sequences in which translational and rotational motion mainly exist.

关 键 词：超分辨率重建 运动估计 最小二乘影像匹配 迭代反投影法 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]