Robustness Evaluation of Local Descriptors
- 10.2991/icaita-16.2016.18How to use a DOI?
- local descriptor; robust; extracting features; image changing
The descriptors should be robust to changes of images taken under various conditions in order to obtain correct recognition. In this paper, we compare the robustness of descriptors computed for local interest regions, for example, extracted by the Harris, SURF and MSER detector. Different descriptors have been compared, depending on the interest region detector. Our evaluation uses different methods of matching features to assess image transformations. We compare SIFT, SURF, SURF with regions detected by MSER, FREAK for interest regions. The results show that local descriptors are sensitive to the viewpoint change, zoom and rotation, and image blur. Illumination has less effect on the robustness of image changing. SURF with regions detected by MSER performs better and is recommended for its better robustness.
- © 2016, the Authors. Published by Atlantis Press.
- Open Access
- This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
Cite this article
TY - CONF AU - Binquan Li AU - Xiaohui Hu PY - 2016/01 DA - 2016/01 TI - Robustness Evaluation of Local Descriptors BT - Proceedings of the 2016 International Conference on Artificial Intelligence: Technologies and Applications PB - Atlantis Press SP - 70 EP - 73 SN - 1951-6851 UR - https://doi.org/10.2991/icaita-16.2016.18 DO - 10.2991/icaita-16.2016.18 ID - Li2016/01 ER -