Contourlet-SIFT特征匹配算法.pdfVIP

第35 卷第5 期 电 子 与 信 息 学 报 Vol.35No.5 2013 年5 月 Journal of Electronics Information Technology May 2013 Contourlet-SIFT 特征匹配算法 陈抒瑢* 李 勃 董 蓉 陈启美 (南京大学电子科学与工程学院 南京 210093) 摘 要：基于局部特征的匹配算法中SIFT(Scale Invariant Feature Transform)算法性能好，应用广泛，但其描述 子的维度高、匹配耗时大，对局部相似区域的匹配鲁棒性差。为此，该文提出一种Contourlet-SIFT 特征匹配算法。 在尺度空间下提取旋转不变特征，对特征及其邻域进行 Contourlet 变换，由各方向子带分解系数的均值和标准差 构建全局纹理描述向量，根据向量间欧氏距离的大小进行特征点排序，选取距离较小的前 1%的特征再进行 SIFT 最近邻比值匹配。实验结果表明该算法对亮度差异大、相似区域多的图像的匹配性能优于SIFT，在保证尺度、旋 转、视角等不变性与SIFT 相当的同时，匹配速度大为提升。 关键词：图像处理；特征匹配；尺度不变特征变换；Contourlet 变换；全局纹理信息 中图分类号：TP391 文献标识码： A 文章编号：1009-5896(2013)05-1215-07 DOI: 10.3724/SP.J.1146.2012.01132 Contourlet-SIFT Feature Matching Algorithm Chen Shu-rong Li Bo Dong Rong Chen Qi-mei (School of Electronic Science and Engineering, Nanjing University, Nanjing 210093, China) Abstract: The Scale Invariant Feature Transform (SIFT) has a fine algorithm performance and an extensive application to the matching algorithm of local features, but its descriptor is characterized by a high dimension and huge time consumption also gives rise to a low matching robustness when tackling similar areas. Therefore this paper puts forward an innovative Contourlet-SIFT feature matching algorithm. The SIFT key points are first extracted to conduct Contourlet transformation on peripheral areas in order to calculate the mean and standard deviation of the decomposition coefficient in each direction. Then the vector of overall texture description is constructed and the Euclidean distance of this low-dimensional vector provides references for prioritizing the matching pairs. The fi