Abstract

Dynamic events are often photographed by a number ofpeople from different viewpoints at different times, result-ing in an unconstrained set of images. Finding the corre-sponding moving features in each of the images allows usto extract information about objects of interest in the scene.Computing correspondence of moving features in such a setof images is considerably more challenging than computingcorrespondence in video due to possible significant differ-ences in viewpoints and inconsistent timing between imagecaptures. The prediction methods used in video for improv-ing robustness and efficiency are not applicable to a set ofstill images. In this paper we propose a novel method topredict locations of an approximately linear moving featurepoint, given a small subset of correspondences and the temporal order of image captures. Our method extends the use of epipolar geometry to divide images into valid and invalid regions, termed Temporal Epipolar Regions (TERs). We formally prove that the location of a feature in a new image is restricted to valid TERs. We demonstrate the effectiveness of our method in reducing the search space for correspon-dence on both synthetic and challenging real world data,and show the improved matching.