A Comparative Study of Various Key-point Detector-Descriptor Algorithms for Augmented Reality Applications
This study aims to compare various feature detector-descriptor algorithms. The algorithms compared are ORB, SIFT, SURF, BRISK, KAZE and AKAZE. The methodology used in this analysis segregates fetching of the frames (from video input) from the actual processing so that time consuming I/O operation does not affect time taken by each algorithm to process the input. This analysis shows that ORB is faster than remaining algorithms with frame processing rate of 23.9 Frames per Second (FPS), while SIFT is more accurate than others for feature detection and description. The results of this analysis can then be used for Augmented Reality applications implementing one of these algorithms.