A Novel Fast and Robust Multimodal Images Matching Method Based on Primary Structure-Weighted Orientation Consistency

The matching of multimodal remote sensing images, especially high-resolution images, is a challenging task due to the nonlinear radiation distortions (NRD), the noise distribution, and the differences in structural texture differences between them. In this article, we proposed a novel fast, robust, and extensible matching method based on the primary structure-weighted orientation consistency (PSOC), which aims to extract relatively consistent primary structures and suppress texture details effectively. To construct the PSOC, we first presented a fast multiscale sigmoid Gabor filter that employs angle interpolation instead of using angle space construction. Then, we enhanced feature representation by using the local primary structure strategy and constructed the PSOC descriptor using an orientation lookup table. Finally, we used the nonmaximum suppressed 3-D normalized cross-correlation fast template matching method for the feature descriptor matching, which improved the matching success rate and reduced the matching complexity under large search radius. In experiments conducted with eight pairs of high-resolution multimodal images, the PSOC descriptor outperformed other state-of-the-art descriptors with an average improvement of 36.6% in correct match rate and an improvement of 22.7% in root mean square error (eliminating the results that these algorithms failed to match). In addition, PSOC achieves efficient matching under large search radius, and the average time complexity is about 1/3 of the other descriptors, which is important for the matching with large offsets in practical applications.

Automated summary

In this article, a novel fast, robust, and extensible matching method based on the primary structure-weighted orientation consistency (PSOC) is proposed to address the challenging task of matching multimodal remote sensing images. The method extracts consistent primary structures and suppresses texture details effectively, leading to improved matching success rate and reduced complexity. Experimental results demonstrate the superiority of the proposed method over state-of-the-art descriptors.