Finding correspondences in wide baseline setups is a challenging problem.Existing approaches have focused largely on developing better featuredescriptors for correspondence and on accurate recovery of epipolar lineconstraints. This paper focuses on the challenging problem of findingcorrespondences once approximate epipolar constraints are given. We introduce anovel method that integrates a deformation model. Specifically, we formulatethe problem as finding the largest number of corresponding points related by abounded distortion map that obeys the given epipolar constraints. We show that,while the set of bounded distortion maps is not convex, the subset of maps thatobey the epipolar line constraints is convex, allowing us to introduce anefficient algorithm for matching. We further utilize a robust cost function formatching and employ majorization-minimization for its optimization. Ourexperiments indicate that our method finds significantly more accurate mapsthan existing approaches.
展开▼
