改进的基于纹理特征的图像配准算法  

Improved image registration algorithm based on texture feature

作　　者：张卫国[1] 李景妹[1] 

机构地区：[1]西安科技大学计算机科学与技术学院,西安710054

出　　处：《计算机工程与应用》2016年第6期214-218,共5页Computer Engineering and Applications

基　　金：国家自然科学基金(No.2012JM8029)

摘　　要：针对图像纹理规则性与重复性及图像起始纹理分布的不一致性,依据VE4000检测系统的纹理特征配准算法达到配准目的。该算法自行设计标准单一模板,对单一模板进行裁剪与拼接技术达到了图像配准的目的,但是配准效果较差,在此基础上提出了改进的纹理特征配准算法,重新设计模板,拼接生成标准的参考图像,对参考图像进行匹配与裁剪以获取不同配准图像。实验结果证明,改进的基于纹理特征的配准算法运行时间短,配准误差小,能够很好满足缺陷检测的配准需求。Aiming at regularity and repeatability of image texture and inconsistent characteristic of initial texture distribution, this paper achieves registration according to the texture feature registration algorithm of VE4000 detection system.This algorithm designs single template by itself, then makes cutting and splicing technology for single template in order to realize the goal of image registration. But the registration effect of this algorithm is relatively poor, so this paper proposes an improved texture feature registration algorithm that designs afresh template, splices for generating standardized referenced image and makes matching and cutting for referenced image to obtain different registration images. Experimental results show that, the improved registration algorithm based on texture feature having shorter running time and smaller registration error, can commendably meet registration requirements of defect detection.

关 键 词：纹理特征 图像配准 模板匹配 图像拼接 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]