基于SIFT和LTP的图像匹配方法

Image matching using SIFT and rotation invariant uniform LTP

针对SIFT算法计算复杂度高,提出了一种SIFT(Scale Invariant Feature Transform)和旋转不变LTP(Local Ternary Pattern)特征相结合的图像匹配方法,以提高SIFT算法的速度.首先利用SIFT算法在两幅需要匹配的图像上分别检测出关键点;然后计算每个关键点周围的旋转不变LTP特征,并作为该关键点的描述子;最后找出两个关键点对之间的匹配点对.实验结果表明,本方法对于图像的匹配性能与SIFT算法相当,运算速度比SIFT算法较快.

In view of high computational complexity of SIFT（scale invariant feature transform）,a new stereo matching algorithm based on SIFT and the rotation-invariant LTP（local ternary pattern） is proposed.Firstly,two sets of key-points are extracted from the two images for matching by utilizing the SIFT algorithm;secondly,the rotation-invariant LTP feature of each key-point is computed,which as the descriptor of the key-point;finally,the matching pairs between the two sets of key-points are determined.The result is compared with the SIFT algorithm in stereo image matching.It is observed that the matching performance by using the SIFT and LTP is the same as the SIFT,and the calculating speed is faster than SIFT.