Homography Matrix-Based Local Motion Consistent Matching for Remote Sensing Images

Feature matching is a fundamental task in the field of image processing, aimed at ensuring correct correspondence between two sets features. Putative matches constructed based on similarity descriptors always contain large number false matches. To eliminate these matches, we propose remote sensing feature method called LMC (local motion consistency), where local consistency refers to property that adjacent have same motion. The core idea find neighborhoods with trends and retain achieve this, design geometric constraint using homography matrix represent consistency. This has projective invariance applicable various types transformations. avoid outliers affecting search for motion, introduce resampling construct neighborhoods. Moreover, jump-out mechanism exit loop without searching all possible cases, thereby reducing runtime. can process over 1000 putative within 100 ms. Experimental evaluations diverse datasets, including SUIRD, RS, DTU, demonstrate achieves higher F-score superior overall performance compared state-of-the-art methods.