A novel algorithm is presented that can track nonplanar templates in a sequence of images.
The algorithm can track features over a larger distance, compared to algorithms that assume planar templates.
The method was incorporated into a visual SLAM algorithm and was shown to improve its accuracy.