Article Details

Abstract

To address the low contrast and strong noise interference of infrared images, which cause insufficient accuracy and uneven feature point distribution in traditional feature matching algorithms, this paper proposes an enhanced infrared image feature matching method. Image quality is improved through multi-scale feature-preserving enhancement preprocessing. A grid homogenization strategy is adopted to optimize the spatial distribution of feature points. The parameters of the scale-invariant feature transform (SIFT) algorithm are adaptively adjusted according to the characteristics of infrared images. Furthermore, an improved fast library for approximate nearest neighbors (FLANN) and random sample consensus (RANSAC) matching strategy is proposed to eliminate mismatches. Experimental results on a self-collected infrared dataset and the open-source DIDLM dataset demonstrate that, compared with the traditional SIFT algorithm, the proposed method reduces the average reprojection error by 31.5% and effectively improves the accuracy and stability of infrared image feature matching. The method can provide reliable support for engineering applications such as target tracking and image stitching.

Keywords

infrared image; feature matching; SIFT algorithm; grid homogenization; feature preservation enhancement

References

1. [1] Chen Shengdi, Yang Li, Zhang Yan, etc Research progress on the application of shortwave infrared technology [J]. Infrared Technology, 2025, 47 (10): 1197-1214
2. [2] Li Yixuan, Li Zhiyang Fast object detection method based on feature matching [J]. Information Technology, 2020, 44 (6): 70-73
3. [3] Liu Hongyu, Li Liheng, Wang Xiaolei Image target scale direction tracking algorithm based on Canny Harris feature points [J]. Information Technology, 2022 (7): 46-50, 58
4. [4] Cui Zhixiang, Lan Chaozhen, Zhang Yongxian, etc A method for matching depth features of thermal infrared and visible light images [J]. Journal of Wuhan University: Information Science Edition, 2023, 48 (2): 316-324
5. [5] Li Ziqian, Baniwamen, Liu Yun, etc A visible light and infrared image matching method based on multi-scale feature point extraction [J]. Infrared Technology, 2025, 47 (3): 351-357
6. [6] Cai Tianwang, Fu Sheng Infrared image registration based on improved SIFT algorithm [J]. Measurement and Control Technology, 2021, 40 (7): 40-45
7. [7] Chen Sijing, Fu Zhitao, Li Ziqian, etc Visible light and infrared image fusion algorithm based on adaptive enhancement and saliency detection [J]. Infrared Technology, 2023, 45 (9): 907-914
8. [8] Hu Junwei, Wang Shiwei, Yang Moyuan Sparse depth feature infrared image stitching algorithm [J]. Infrared Technology, 2025 (5): 584-590
9. [9] Zhang D，Liu Y N，Zhao Y，et al. Algorithm Research on Detail and Contrast Enhancement of High Dynamic Infrared Images[J/OL].Applied Sciences，2023，13（23）：12649.https://doi.org/10.3390/app132312649.
10. [10] Gong W S，He C，Su K J，et al. DIDLM: A SLAM Dataset for Difficult Scenarios Featuring Infrared, Depth Cameras, LiDAR, 4D Radar, and Others Under Adverse Weather, Low Light Conditions, and Rough Roads[J].IEEE Transactions on Intelligent Transportation Systems，2025，27（2）：1908-1920.

About the authors

Gong Xiaoqiang (November 1972-), male, Han ethnicity, from Xi'an, Shaanxi, intermediate professional title, bachelor's degree, research interests include multimodal information fusion and image processing.