一种基于LTP特征的图像匹配方法  

Image matching method based on LTP features

作　　者：赵灵芝[1] 李伟生[1] 

机构地区：[1]重庆邮电大学计算机科学与技术研究所,重庆400065

出　　处：《计算机应用研究》2009年第10期3983-3985,共3页Application Research of Computers

基　　金：重庆市自然科学基金资助项目(CSTC2006BB2365);重庆市教育委员会科学技术研究资助项目(KJ060504)

摘　　要：为了解决可变光照条件下图像匹配困难这一问题,提出一种基于LTP(局部三值模式)特征的图像匹配方法。LTP是LBP(局部二值模式)的扩展,这种局部纹理描述算法较LBP更具有判别能力而且对于统一区域的噪声更不敏感。利用LTP对于图像旋转和光照变化都具有良好的鲁棒性,解决了在光照可变条件下流行的SIFT方法进行匹配的困难。通过对不同变换的图像进行匹配实验表明,该方法得到的匹配结果比LBP效果更好,鲁棒性更高,而在光照可变和噪声很大的情况下比流行的SIFT方法更实用。Due to solve the problem that image matching was very difficult when illumination condition was changeable, proposed an image matching method based on LTP （ local ternary patterns） features of image. Local ternary patterns was a generalization of the local binary pattern （LBP） and its local texture descriptor was more discriminate and less sensitive to noise in uniform regions. Because LTP feature was very robust to illumination change and rotation of image, it could be used to solve the difficulty that current popular SIFT method was not capable of matching two images under changeable illumination condition. It experimented on a large of transformed images, the results show that the result of this method is more efficient and robust, but when the illumination is changeable the results of method are more effective than SIFT.

关 键 词：图像匹配 局部三值模式 局部二值模式 尺度不变特征变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]