Research on the Fundamental Principles and Characteristics of Correspondence Function

摘要: The correspondence function (CF) is a concept recently introduced to reject the mismatches from given putative correspondences. fundamental idea of CF is that relationship some corresponding points between two images be registered can described by pair vector-valued functions, estimated nonparametric regression method with more flexibility than normal parametric model, for example, homography matrix, similarity transformation, and projective transformations. Mismatches are rejected by checking their consistency CF. This paper proposes a visual scheme investigate principles the CF and studies its characteristics experimentally comparing it widely used model epipolar geometry (EG). It shown that describes mapping the points in one image another image, which enables direct estimation positions the corresponding points. In contrast, EG acts reducing the search space for two-dimensional space line, problem one-dimensional space. As a result, undetected usually near the correct points, but many undetected mismatches far point.