3D Metric Reconstruction from Uncalibrated Omnidirectional Images

We show that it is possible to obtain a very complete 3D metric reconstruction of the surrounding scene from two or more uncalibrated omnidirectional images. In particular, we demonstrate that omnidirectional images with angle of view above 180 can be reliably autocalibrated. We also show that wide angle images provide reliable information about their camera positions and orientations. We link together a method for simultaneous omnidirectional camera model and epipolar geometry estimation and a method for factorization-based 3D reconstruction in order to obtain metric reconstruction of unknown scene observed by uncalibrated omnidirectional images. The 3D reconstruction is done from automatically established image correspondences only. We demonstrate our method in experiments with Nikon FC–E8 and Sigma 8mm-f4-EX fish-eye lenses. Nevertheless, the proposed method can be used for a large class of non-perspective central omnidirectional cameras.

[1]  P. Sturm.  Mixing catadioptric and perspective cameras , 2002, Proceedings of the IEEE Workshop on Omnidirectional Vision 2002. Held in conjunction with ECCV'02.

[2]  Rachid Deriche,et al.  A Robust Technique for Matching two Uncalibrated Images Through the Recovery of the Unknown Epipolar Geometry , 1995, Artif. Intell..

[3]  Ben J. A. Kröse,et al.  Robust scene reconstruction from an omnidirectional vision system , 2003, IEEE Trans. Robotics Autom..

[4]  Shree K. Nayar,et al.  Ego-motion and omnidirectional cameras , 1998, Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271).

[5]  Peter F. Sturm,et al.  A Factorization Based Algorithm for Multi-Image Projective Structure and Motion , 1996, ECCV.

[6]  Kostas Daniilidis,et al.  Paracatadioptric Camera Calibration , 2002, IEEE Trans. Pattern Anal. Mach. Intell..

[7]  Tomáš Svoboda,et al.  Reliable 3D reconstruction from a few catadioptric images , 2002, Proceedings of the IEEE Workshop on Omnidirectional Vision 2002. Held in conjunction with ECCV'02.

[8]  Kostas Daniilidis,et al.  Structure and motion from uncalibrated catadioptric views , 2001, Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. CVPR 2001.

[9]  Tomás Pajdla,et al.  Omnidirectional Camera Model and Epipolar Geometry Estimation by RANSAC with Bucketing , 2003, SCIA.

[10]  Richard Szeliski,et al.  3-D Scene Data Recovery Using Omnidirectional Multibaseline Stereo , 1996, Proceedings CVPR IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[11]  M. Hebert,et al.  Omni-directional structure from motion , 2000, Proceedings IEEE Workshop on Omnidirectional Vision (Cat. No.PR00704).

[12]  Karl Meerbergen,et al.  The Quadratic Eigenvalue Problem , 2001, SIAM Rev..

[13]  Sing Bing Kang,et al.  Catadioptric self-calibration , 2000, Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).

[14]  Ryad Benosman,et al.  Calibration of panoramic catadioptric sensors made easier , 2002, Proceedings of the IEEE Workshop on Omnidirectional Vision 2002. Held in conjunction with ECCV'02.

[15]  J. P. Mellor.  Geometry and Texture from Thousands of Images , 2004, International Journal of Computer Vision.

[16]  Jiri Matas,et al.  Robust wide-baseline stereo from maximally stable extremal regions , 2004, Image Vis. Comput..

[17]  Tomás Pajdla,et al.  Estimation of omnidirectional camera model from epipolar geometry , 2003, 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings..

[18]  Tomás Pajdla,et al.  Structure from Many Perspective Images with Occlusions , 2002, ECCV.

[19]  David W. Jacobs,et al.  Linear fitting with missing data: applications to structure-from-motion and to characterizing intensity images , 1997, Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[20]  Bernhard P. Wrobel,et al.  Multiple View Geometry in Computer Vision , 2001 .