Image feature points matching via improved ORB

An improved ORB (Oriented FAST and Rotated BRIEF) algorithm motivated by SIFT (Scale-Invariant Feature Transform) is put forward aimed at solving the deficiency that ORB has little scale invariance for feature points matching. Firstly, the scale spaces were built for the detection of stable extreme points, and the stable extreme points detected were considered to be feature points with scale invariance. Secondly, ORB descriptor was used to describe the feature points, which would finally form the binary descriptors with scale and rotation invariance. Finally, point matching was conducted according to Hamming distance. Experimental results show that the proposed algorithm achieves good matching performance in terms with scale invariance taking into consideration. When images have scale changes, feature points matching accuracy rate of the improved algorithm can reach with about 92.53%, which improves the matching accuracy rate by about 68.27% compared to ORB. In addition, the matching speed of the improved algorithm, which inherits the fast superiority of ORB, is about 65.28 times faster than SIFT averagely.