Image matching for structured scenes based on ASIFT and homography constraint
You Zhai, Guangjun Yu, Hongyun Wang, Xiwei Guo · 2017
Images of structured scenes have uniform texture and few matches can be gotten in wide baseline case. Therefore, matching method based on ASIFT and homography constraint is presented. First, initial matches are obtained using ASIFT and epipolar geometry constraint. Second, the initial matches are divided into different groups based on homography constraints and homography matrix is estimated using RANSAC based strategy for each group. Third, the FAST features and homography constraints are used to extend the initial matches. Finally, homography constraints are updated using extended matches and outliers are removed with epipolar geometry constraint. Experiment results show that the proposed method can adapt to wide base image matching for structured scenes images matching and more matches can be obtained compared to ASIFT, which lay good foundation for accurate epipolar geometry estimation and three dimensional reconstruction.