Camera self-calibration from ellipse correspondences

In this paper, we introduce a new technique for camera self-calibration using ellipse correspondences. Based on an analysis of ellipse matches between the images obtained from the same viewpoint but with different and unknown view directions, our approach estimates the intrinsic camera parameters. We present both liner and non-linear solutions to recovering intrinsic camera parameters. The algorithm's performance is validated extensively using both synthetic and real image data. Compared with similar techniques but using points, we observe a comparable performance. The use of ellipses, however, greatly simplifies feature matching between images, improves matching accuracy, and avoids mismatch.