Comparative analysis of the feature extraction performance of augmented reality algorithms

Abstract

The algorithms that extract keypoints and descriptors in augmented reality applications are getting more and more important in terms of performance. Criterions like time and correct matching of points gain more impact according to the type of application. In this paper, the performance of the algorithms used to identify an image using keypoint and descriptor extraction is studied. In the context of this research, main criterion like the number of keypoints and descriptors that the algorithms extract, algorithm execution time, and the quality of keypoints and descriptors extracted are considered as the performance metrics. Same data stacks were used for obtaining comparison results. In addition to comparisons for a group of well-known augmented reality applications, the best performing algorithms for varying applications were also suggested. C++ language and OpenCV library were used for the implementation of the augmented reality algorithms compared.