Keypoint Matches Filtering in Computer Vision: Comparative Analysis of RANSAC and USAC Variants

In this study, a detailed analysis is conducted to evaluate the efficiency of various keypoint matching filtering methods, including RANSAC and its USAC variations, namely, USAC-DEFAULT, USAC-FAST, USAC-ACCURATE, USAC-MAGSAC, and USAC-PROSAC. Keypoints are detected and described using the SIFT, SURF, ORB, and BRISK methods. This work aims to assess the impact of filtering methods on the accuracy, stability, and processing speed of image analysis. The results show that while RANSAC has the slowest performance, it provides the highest stability, with a similarity coefficient deviation of 0.5%. RANSAC with modified parameters demonstrates higher accuracy and significantly faster processing compared to standard RANSAC, outperforming it by approximately 2.5 times and achieving a 4% accuracy improvement over USAC-DEFAULT. The most rapid methods are USAC-PROSAC and USAC-FAST, whereas USAC-MAGSAC has the longest execution time among all USAC variations. Accuracy analysis of the different detectors shows that SIFT achieved the highest similarity coefficient values. SURF demonstrated slightly lower accuracy than SIFT, while BRISK showed results inferior to SURF. ORB is found to be the least effective among the evaluated detectors. This work emphasizes the importance of an adaptive approach when selecting keypoints matching filtering methods to achieve high accuracy, stability, and processing speed in various computer vision applications. The findings of this study will assist developers and researchers in choosing optimal filtering methods and improving the efficiency of image processing algorithms for specific tasks.