图像校准中的多元样条方法  

Image registration by using multi-variate spline functions

作　　者：郑鹏 郑可妮 刘凤山 施锡泉 

机构地区：[1]Department of Mathematical Sciences, Delaware State University

出　　处：《中国科学：数学》2015年第9期1599-1610,共12页Scientia Sinica：Mathematica

摘　　要：图像校准是将两两具有公共部分的一组图像通过变换合成到同一个图像的过程.通常这组图像可以是从不同时间、不同传感器(成像设备)或不同条件(气候、亮度、摄像位置和角度等)下获取的.基于任意三角剖分上的多元样条函数,本文提出了图像配准的一种新方法.实验结果表明,与目前国际上流行的其他图像校准方法相比,本文方法计算简单并且得到的合成图像具有更为自然的边界过渡.Image registration is a kind of image processing that merges a series of pictures into a global picture. Any picture of this series has a common part with another one and they may be produced in different time, by different equipments, and under different factors such as weather, illumination angle etc. Based on multivariate splines defined on arbitrary triangulations, in this paper a new image registration method is presented. Experiment results show that, comparing with other methods, our results have better merging texture with simpler and easier calculation.

关 键 词：图像校准 多元样条 三角剖分 仿射变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]