移动平台上四邻域梯度阈值的 SIFT 匹配优化

摘要

To improve the robustness and efficiency of the matching algorithm based on SIFT (scale invariant feature transform),a grouping matching method based on gradient value of local neighborhood was proposed,making it possible to use SIFT on mobile devices such as instruments.A new feature of the detected points was identified according to the maximum gradient value of the four neighbor-hood points around,which was used to group the key-points before matching.The performance of the proposed methods was tested on groups of images,and compared with the traditional algorithm.The experimental results show that the feature matching speed can be ac-celerated about 2 to 3 times using the proposed method without losing a noticeable amount of correct matches.%为提高基于 SIFT （scale invariant feature transform）图像匹配算法的匹配鲁棒性和效率，以便在仪器仪表等便携设备上使用，提出一种基于邻域梯度值分组的匹配方法。将特征点周围四邻域的最大梯度值作为特征之一，在 SIFT 特征匹配之前作为分组匹配依据。在 Android 移动平台上，对优化算法和传统算法进行对比实验，实验结果表明，改进后的SIFT 算法能够在保证匹配正确率的同时减少运算量，匹配速度可提高2至3倍。