一种基于Harris角点特征精确匹配的图像拼接方法  

An Image Mosaic Method Based on Refined Corner Points Feature Matching

作　　者：涂春萍[1] 柴亚辉[1] 李广丽[1] 刘觉夫[1] 

机构地区：[1]华东交通大学信息工程学院,江西南昌330013

出　　处：《实验室研究与探索》2011年第10期40-43,共4页Research and Exploration In Laboratory

摘　　要：提出了一种基于Harris角点特征的精确匹配方法。该方法首先提取参考图像及待拼接图像中各自的Harris角点点集,并计算出这2个点集间每对点的圆形邻域图像的相关系数;再通过提取各个角点邻域的Hu矩特征,获得了该特征下每对点的相似程度。将不同特征下的相似度进行归一化并融合,构造出2个点集间,每对点的相似度表。在此表的基础上,优化匹配结果,使得匹配点对的总体相似程度高,从而得到精确匹配。由于Hu矩特征具有旋转及尺度不变性,因此提取出的角点特征能够较好地抵抗常见的图像变换。最后,实现了一套包括图像预处理、图像对齐与匹配等诸多模块的图像拼接系统。通过实际操作表明,该方法的图像拼接效率较高,有较好的鲁棒性。In this paper, a refined feature matching approach based on Harris Corners Detection was proposed. This approach first extracts the Harris corner point sets within the referring image and the image to-be-processed. Then, it calculates the correlation between each corner point from both point sets within a circular area around it. After that, it extracts the feature of Hu moment from each point,and attains the similarity. By normalization and fusion of the similarity under different features,the similarity table is constructed,based on which ,the matching is optimized. Due to the rotation and scaling invariance of Hu moment,the very feature is resistant to the common transformation of the image. Finally,an Image Mosaic System was realized,which consists of preprocessing and feature matching of the images. Practical operation proves that the algorithm can provide an efficient and robust process of image mosaic.

关 键 词：图像拼接 HARRIS角点检测 HU矩 特征匹配 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]