COMPREHENSIVE SPATIAL-GEOMETRIC EVALUATION OF KEYPOINT DETECTORS

Andriy Fesiuk, Yuriy Furgala

Abstract


Background. Local features are essential components of modern computer vision systems, such as SLAM and 3D reconstruction. Traditional evaluation protocols for keypoint detection mainly focus on geometric accuracy and repeatability, often neglecting the spatial structure of the point distribution. This complicates algorithm selection for applications where uniform image coverage and the absence of excessive local clustering are important. This work aims to conduct a comprehensive comparison of keypoint detectors using an extended set of metrics that account for both geometric accuracy and the spatial properties of features.

Materials and Methods. The study was conducted on the HPatches dataset using six detectors: SIFT, SURF, ORB, BRISK, KAZE, and AKAZE. Keypoint filtering and geometric verification of correspondences were performed using USAC. Matching quality was assessed through the geometric metrics MMA, Repeatability, and Verification Ratio. Spatial analysis used the metrics CUI, RI, and SCS. To compare keypoint detection methods, a quality index Q was introduced that integrates geometric and spatial indicators.

Results and Discussion. The study showed that selecting points by response strength significantly improves matching accuracy for SIFT, ORB, and BRISK, but may lead to local redundancy of keypoints. KAZE and AKAZE demonstrated the best overall balance, achieving high accuracy along with more uniform scene coverage. ORB tended to form dense clusters in high-contrast regions, thereby reducing its structural effectiveness, whereas SURF consistently delivered high performance regardless of the keypoint selection strategy.

Conclusion. The proposed evaluation method allows a consistent analysis of the geometric and spatial properties of keypoint detectors. It shows that, for a fixed number of keypoints, the performance of the final method depends not only on the geometric accuracy of matches but also on the features of the spatial point distribution. It was observed that the keypoint selection process, especially response-based selection, systematically affects both geometric and spatial characteristics. The Q quality index combines these aspects into a single metric. It can be used to compare detection methods in scenarios that require both reliable matches and well-balanced scene coverage.

Keywords: feature detection, spatial distribution, geometric metrics, image matching.


Full Text:

PDF

References


[1] Zhou, L., Wu, G., Zuo, Y., Chen, X., & Hu, H. (2024). A comprehensive review of vision-based 3D reconstruction methods. Sensors, 24(7), 2314. https://doi.org/10.3390/s24072314

[2] Ye, Z., Bao, C., Zhou, X., Liu, H., Bao, H., & Zhang, G. (2023). EC-SfM: Efficient covisibility-based structure-from-motion for both sequential and unordered images. IEEE Transactions on Circuits and Systems for Video Technology. https://doi.org/10.1109/TCSVT.2023.3285479

[3] Huang, Q., Guo, X., Wang, Y., Sun, H., & Yang, L. (2024). A survey of feature matching methods. IET Image Processing, 18(6), 1385-1410. https://doi.org/10.1049/ipr2.13032

[4] Herrera-Granda, E. P., Berrones-González, A., & Aguilar, W. (2024). Monocular visual SLAM, visual odometry, and structure from motion: A review. Heliyon, 10(9), e37356. https://doi.org/10.1016/j.heliyon.2024.e37356

[5] Al-Tawil, Y., Zagraoui, A., & Zeghbib, A.-R. (2024). A review of visual SLAM for robotics: evolution, properties, and future applications. Frontiers in Robotics and AI, 11, 1347985. https://doi.org/10.3389/frobt.2024.1347985

[6] Ehsan, S., Kanwal, N., Clark, A. F., & McDonald-Maier, K. D. (2011). Measuring the coverage of interest point detectors. In M. Kamel & A. Campilho (Eds.), Image Analysis and Recognition (ICIAR 2011) (Lecture Notes in Computer Science, Vol. 6753, pp. 253-261). Springer. https://doi.org/10.1007/978-3-642-21593-3_26

[7] Mousavi, V., Varshosaz, M., & Remondino, F. (2021). Using information content to select keypoints for UAV image matching. Remote Sensing, 13(7), 1302. https://www.mdpi.com/2072-4292/13/7/1302

[8] Balntas, V., Lenc, K., Vedaldi, A., & Mikolajczyk, K. (2017). HPatches: A benchmark and evaluation of handcrafted and learned local descriptors. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). https://doi.org/10.1109/CVPR.2017.410

[9] Balntas, V., Lenc, K., Vedaldi, A., Tuytelaars, T., Matas, J., & Mikolajczyk, K. (2020). H-Patches: A benchmark and evaluation of handcrafted and learned local descriptors. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42(11), 2825-2841. https://doi.org/10.1109/TPAMI.2019.2915233

[10] Lowe, D. G. (2004). Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 60(2), 91-110. https://doi.org/10.1023/B:VISI.0000029664.99615.94

[11] Bay, H., Ess, A., Tuytelaars, T., & Van Gool, L. (2008). Speeded-Up Robust Features (SURF). Computer Vision and Image Understanding, 110(3), 346-359. https://doi.org/10.1016/j.cviu.2007.09.014

[12] Alcantarilla, P. F., Bartoli, A., & Davison, A. J. (2012). KAZE features. In ECCV 2012 (LNCS 7577, pp. 214-227). https://doi.org/10.1007/978-3-642-33783-3_16

[13] Alcantarilla, P. F., Nuevo, J., & Bartoli, A. (2013). Fast explicit diffusion for accelerated features in nonlinear scale spaces. In BMVC 2013 (pp. 1-11). https://doi.org/10.5244/C.27.13

[14] Rublee, E., Rabaud, V., Konolige, K., & Bradski, G. (2011). ORB: An efficient alternative to SIFT or SURF. In ICCV 2011 (pp. 2564-2571). https://doi.org/10.1109/ICCV.2011.6126544

[15] Leutenegger, S., Chli, M., & Siegwart, R. Y. (2011). BRISK: Binary robust invariant scalable keypoints. In ICCV 2011 (pp. 2548-2555). https://doi.org/10.1109/ICCV.2011.6126542

[16] M. Ivashechkin, D. Baráth, J. Matas, "USACv20: Robust Essential, Fundamental and Homography Matrix Estimation," 2021. https://doi.org/10.48550/arXiv.2104.05044

[17] Howse, Joseph, and Joe Minichino. “Learning OpenCV 4 Computer Vision with Python 3: Get to grips with tools, techniques, and algorithms for computer vision and machine learning.”, Packt Publishing Ltd, 2020.

[18] Fesiuk, A., & Furgala, Y. (2025). Keypoint matches filtering in computer vision: Comparative analysis of RANSAC and USAC variants. International Journal of Computing, 24(2), 343-350. https://doi.org/10.47839/ijc.24.2.4018

[19] Hossein-Nejad, Z., & Nasri, M. (2017). RKEM: Redundant keypoint elimination method in image registration. IET Image Processing, 11, 273-284. https://doi.org/10.1049/iet-ipr.2016.0440

[20] Hossein-Nejad, Z., Agahi, H., & Mahmoodzadeh, A. (2021). Image matching based on the adaptive redundant keypoint elimination method in the SIFT algorithm. Pattern Analysis and Applications, 24(2), 669-683.

https://doi.org/10.1007/s10044-020-00938-w

[21] Hossein-Nejad, Z., & Nasri, M. (2022). Clustered redundant keypoint elimination method for image mosaicing using a new Gaussian-weighted blending algorithm. The Visual Computer, 38(6), 1991-2007.

https://doi.org/10.1007/s00371-021-02261-9

[22] Harris, C., & Stephens, M. (1988). A combined corner and edge detector. In Proceedings of the 4th Alvey Vision Conference (pp. 147-151). https://doi.org/10.5244/C.2.23

[23] Canny, J. (1986). A computational approach to edge detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 8(6), 679-698. https://doi.org/10.1109/TPAMI.1986.4767851

[24] Soille, P. (2004). Morphological Image Analysis: Principles and Applications (2nd ed.). Springer. https://doi.org/10.1007/978-3-662-05088-0

[25] Rey-Otero, I., Delbracio, M., & Morel, J.-M. (2015). Comparing feature detectors: A bias in the repeatability criteria. In 2015 IEEE International Conference on Image Processing (ICIP) (pp. 3024-3028). https://doi.org/10.1109/ICIP.2015.7351358

[26] Rey-Otero, I., & Delbracio, M. (2015). Is repeatability an unbiased criterion for ranking feature detectors? SIAM Journal on Imaging Sciences, 8(4), 2558-2580. https://doi.org/10.1137/15M1007732

[27] Avriel, M., & Wilde, D. J. (1968). Golden block search for the maximum of unimodal functions. Management Science, 14(5), 307-319 https://doi.org/10.1287/mnsc.14.5.307

[28] Boudt, K., d’Errico, M., Luu, H. A., & Pietrelli, R. (2022). Interpretability of Composite Indicators Based on Principal Components. Journal of Probability and Statistics, 2022, 4155384, 1-12. https://doi.org/10.1155/2022/4155384

[29] Mukherjee, S., Lagache, T., & Olivo-Marin, J.-C. (2021). Evaluating the stability of spatial keypoints via cluster core correspondence index. IEEE Transactions on Image Processing, 30, 386-401. https://doi.org/10.1109/TIP.2020.3036759

[30] A. Fesiuk and Y. Furgala, (2025). Comparative Study of Feature Detectors and Keypoint Filters in Image Matching. Electronics and Information Technologies, (31), (pp. 71-88). http://dx.doi.org/10.30970/eli.31.7




DOI: http://dx.doi.org/10.30970/eli.32.5

Refbacks

  • There are currently no refbacks.