SIFT算法的相似性度量优化

摘要

针对人脸姿势变化对人脸识别的影响,采用具有良好的尺度、旋转、光照等不变特性的SIFT算法.在分析SIFT特征向量生成过程的基础上,提出一种降低SIFT匹配过程中相似性度量计算代价的方法以提高SIFT特征匹配效率.该方法以绝对值距离和棋盘距离的线性组合距离代替欧式距离作为特征描述子之间的相似性度量.实验结果表明:该方法在保持鲁棒性的同时,可以降低时间复杂度,提高图像匹配的效率.%According to the affection of the change of the face position on the face recognition,we adopt the algorithm of Scale Invariant Feature Transformation that has the good invariance of scale,rotation,and illumination.To improve SIFT feature matching algorithm efficiency,a method of reducing similar measure matching cost is presented.The linear combination of cityblock distance and chessboard distance is employed in place of Euclidean distance,and it achieves comparability measurement of feature descriptors.The experimental results show that the algorithm can reduce the rate of time complexity and maintain robust quality at the same time,and the image matching efficiency is improved.