RI-LPOH: Rotation-Invariant Local Phase Orientation Histogram for Multi-Modal Image Matching

To better cope with the significant nonlinear radiation distortions (NRD) and severe rotational in multi-modal remote sensing image matching, this paper introduces a rotationally robust feature-matching method based on maximum index map (MIM) 2D matrix, which is called rotation-invariant local phase orientation histogram (RI-LPOH). First, feature detection performed weighted moment equation. Then, matrix MIM modified gradient location (GLOH) constructed invariance achieved by cyclic shifting both column row directions without estimating principal separately. Each part of sensed image’s additionally flipped up down to obtain another avoid intensity inversion, all matrices are concatenated rows form final 1D vector. Finally, RFM-LC algorithm introduced screen obtained initial matches reduce negative effect caused high proportion outliers. On basis, remaining outliers removed fast sample consensus (FSC) optimal transformation parameters. We validate RI-LPOH six different types datasets compare it four state-of-the-art methods: PSO-SIFT, MS-HLMO, CoFSM, RI-ALGH. The experimental results show that our proposed has obvious advantages success rate (SR) number correct (NCM). Compared RI-ALGH, mean SR 170.3%, 279.8%, 81.6%, 25.4% higher, respectively, NCM 13.27, 20.14, 1.39, 2.42 times aforementioned methods.