基于偏最小二乘的SIFT误匹配校正方法

针对尺度不变特征变换（SIFT）描述子仅利用特征点的局部邻域信息而对图像内具有相似结构的特征点易产生误匹配的现象，提出一种基于偏最小二乘的SIFT误匹配校正方法。该方法首先利用SIFT算法进行匹配，得到初始匹配对，然后利用偏最小二乘方法对匹配后初始匹配点的空间分布信息进行重新描述，并通过定义影响函数，剔除影响程度大的特征点对，最后得到精确匹配点对，对图像进行配准。实验结果表明，该方法能够有效地剔除误匹配点，提高图像配准的精度。

SIFT mismatching correction based on partial least squares

A method to correct Scale Invariant Feature Transform(SIFT) mismatching based on Partial Least Squares(PLS) was presented to solve the mismatching when the feature points locate in some similar structures of one image.At first,SIFT algorithm was used to extract the feature points,and get initial matching.Then,each matching point obtained by SIFT could be described again using PLS.Furthermore,the introduction of influence function helped to judge and remove the mismatching.The experimental results on optical and remote sensing images show that the proposed algorithm outperforms conventional approach in terms of mismatching correction.