基于改进多尺度ORB的图像配准算法

An Image Registration Algorithm Based on ImprovedMulti-scale ORB

针对现有多尺度ORB算法精确度不高、运算时间长的问题,提出一种改进算法。首先以下采样加滤波的方法建立多尺度空间提取ORB特征点;采用最近邻法匹配特征点,利用匹配点邻域灰度分布相似性,将特征点的显著性引入余弦相似度筛选匹配点;结合图像的局部熵改进了K-means算法,只在特征点密集区域内检测初始类中心点,使特征点根据图像内容更合理地分类;再使用高效的改进PROSAC算法计算图像间的变换矩阵,在保证变换矩阵准确性的情况下,进一步减少运算量;最后通过两组实验验证了算法性能。

For the existing multi-seale ORB algorithms having low accuracy and long operation time,animproved algorithm is proposed.Firstly,a multi-scale space is established by down sampling and filtering toextract ORB feature points,the nearest neighbor method is used to match feature points,and the significanceof feature points is introduced into cosine similarity to sereen matching points by using the similarity of graydistribution in the neighborhood of matching points.Combined with the loeal entropy of the image,theK-means algorithm is improved,and the initial class center point is detected only in the area with densefeature points,so that the feature points can be classified more reasonably according to the image.Then,theimproved PROSAC algorithm is used to calculate the transformation matix between images,which futherreduces the computational complexity while ensuring the accuracy of the transformation matrix.Finally,theperformance of the algorithm is verified by two groups of experiments.