Optical-to-SAR Image Matching Based on Fast Rank-Based Local Self-Similarity Descriptor
Xin Xiong, Guowang Jin, Ruibing Cui, Miao Xu, He Yang · 2025
Automatic optical-to-synthetic aperture radar (SAR) image matching remains a challenging task due to significant radiometric differences between the two modalities. To address this challenge, we propose a robust optical-to-SAR image template matching method based on a fast rank-based local self-similarity (FRLSS) descriptor. The proposed FRLSS descriptor is an accelerated variant of our previously introduced rank-based local self-similarity (RLSS) descriptor, achieved through the application of an offset mean filtering (OMF) method. Furthermore, we optimized the structure of the FRLSS descriptor and integrated it into the dense scheme to derive the DFRLSS descriptor, thereby enhancing its matching performance. Experimental results from nine optical-to-SAR image pairs demonstrate that the proposed method surpasses three state-of-the-art methods, exhibiting superior robustness in both matching accuracy and computational efficiency.