Comparative Research on Correspondence Function

Correspondence Function (CF) is a recently introduced concept for rejecting outliers in image point matching. The fundamental idea of CF is that, for given two images I´ and I´ of a scene, the relationships between the corresponding points can be described by a pair of functions f:I->;I´ and f´:I´->;I. And the outliers can be detected by checking whether they are consistent with the Correspondence Function. In this work, we compared the CF with the most popular model Fundamental matrix theoretically and experimentally. Experimental results on typical image pairs demonstrate its superiority over the available methods in literature, especially the widely used RANSAC. In addition, the ICF is a general method for rejecting mismatches, and it is applicable to images both of rigid objects and of nonrigid objects with unknown deformation.