Stereo Error Detection, Correction, and Evaluation

Even with an increased sophistication of stereo matching techniques, only incremental gains have been made in their perfor- mance. The purpose of this paper is to present an algorithm for error detection and correction of disparity, as a process separate from stereo matching, with the contention that matching is not necessarily the best place to exploit all the physical constraints characteristic to stereopsis. As a result of the bias in stereo research towards matching, vision tasks like surface interpolation and object modeling have to accept erro- neous data from the stereo matchers without the benefit of any inter- vening stage of error correction. We present an algorithm which iden- tifies all errors in disparity data which can be detected on the basis of figural continuity and corrects them. The algorithm can he used as a post processor to any edgel based stereo matching algorithm' and can additionally be employed to automatically provide quantitative evalu- ations of the performance of matching algorithms of this class. Index Terms-Linear features, matching, stereo.