基于改进SIFT特征匹配的快速图像拼接算法

A Fast Image Mosaic Algorithm Based on Improved SIFT Feature Matching

针对经典SIFT算法存在特征点冗余、运算量大、鲁棒性差的问题,提出一种快速、准确的图像拼接算法.在SIFT提取不变特征向量过程中加入多尺度Harris检测算子,筛选出更能代表图像信息的特征点.利用快速RANSAC算法(PERANSANC)计算图像之间的变换矩阵.把检测到的特征点作为图像中的运动目标,运用KLT跟踪算法,确定特征点的精确位置,精炼变换矩阵.实验验证了该算法的稳健性和快速性.

Considering the redundant feature points, high computational complexity, low robustness in image mosam method based on scale invariant feature transform （SIFT） algorithm, a steady and fast image mosaic algorithm is proposed. This paper introduces Harris corner operator into the process of extracting feature points based on SIFT algorithm. The improved SIFT algorithm picks out Harris corners from the SIFT feature points when checking them in multi-- scale space, chooses feature points that stand for image information. PERANSANC algorithm is used to compute the homography matrix between the two images. Then the detected corners are regarded as moving objects in the images, located accurately by the KLT track algorithm and the homography matrix is refined. The experiments verify the algorithm＇s efficiency and robustness.