一种结合SIFT和对应尺度LTP综合特征的图像匹配算法  

An image matching algorithm combining SIFT and its corresponding scale LTP features

作　　者：陈丽芳[1] 刘一鸣[1] 刘渊[1] 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122

出　　处：《计算机工程与科学》2015年第3期582-588,共7页Computer Engineering & Science

基　　金：江苏省基金重点资助项目(BK2011003);江苏省自然科学基金青年基金资助项目(BK20130161)

摘　　要：由于SIFT特征是一种性能良好的局部特征,常被广泛应用于图像匹配,但SIFT特征点有128维描述符,所以具有匹配复杂度高和计算量大等缺点。为了提高图像匹配效率,研究了一种新的图像匹配方法。该方法通过构建尺度空间、检测极值点、确定关键点等步骤生成SIFT关键特征点;然后利用特征点周围邻域点的旋转不变LTP特征和相对灰度直方图来描述,替代传统SIFT特征点的128维描述,图像匹配过程中使用街区距离代替欧氏距离;最后利用光照变化、模糊变化、尺度和旋转综合变化三组图像进行算法仿真匹配实验。实验结果表明,本算法在图像尺度、旋转、光照变化条件下具有更高的匹配精确度,并且有效地提高了图像的匹配速度。SIFT is one of the most robust and widely used image matching algorithms based on local features.However,the computational and matching complexity of the descriptor with 128 dimensions is too high.In order to speed up image matching,an image matching algorithm is proposed.Firstly,measures such as constructing scale spaces,detecting extrema and confirming key-points are carried out to create the key feature points.Secondly,the key feature points are described by the rotation-invariant LTP which is formed by the patches around the key-points,and by the relative grayscale histogram.Lastly,rotation invariant descriptors are formed,and the city-block distance is adopted for image matching.Experimental results prove that the robust method improves the matching speed and matching precision even under the condition of illumination transformation,scale transformation,and rotation transformation.

关 键 词：图像匹配 SIFT 对应尺度LTP 相对灰度直方图 街区距离 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]