Learning Correspondence View with Support Vector Machine

Correspondence view (CV) is recently introduced for rejecting outliers in computer vision. The fundamental idea of CV is that, for given two images of a scene, the corresponding points constitute a manifold in joint-image space , and outliers can be detected by checking whether they are consistent with the upward views of the manifold. This work studies CV learning and outliers rejecting by support vector machine. Experiments on real image pairs demonstrate the excellent performance of our proposed SVM+CV learning method and its superiority over the available robust methods in literature, especially the widely used RANSAC.