Skip to content

3D Reconstruction by Gluing Pair-wise Euclidean Reconstructions

or   “How to Achieve a Good Reconstruction from Bad Images” 3DPVT’06 Demo

Daniel Martinec, Tomáš Pajdla, Jana Kostková, and Radim Šára

Given a set of unorganized images taken from unknown viewpoints and directions, the proposed method [1] estimates camera positions. Such automatic calibration makes it possible to compute an accurate 3D model of the object. The only assumption is a sufficient overlap between some image pairs.

Animations will be seen, e.g.,
in Cortona VRML Client
under Windows and freewrl
under Linux.


26 images [1136×852]
distinguished regions (DR)
91.21% occlusions

23 img [1413×942]
96.88% occlusions
Detenice fountain

54 img [1200×900]
98.43% occlusions
Daliborka tower

64 images [1100×850]
96.02% occlusions
ICCV’05 Contest finals

38 images [1120×840]
94.05% occlusions
St. Martin rotunda

104 images [972×1296]
96.55% occlusions

Reconstructions of the sparse correspondences were obtained using method [1] and the dense reconstructions using methods [2], [3] and [4].

Fully automatic data processing pipeline [4]: Highly discriminative features are first detected in all images. Correspondences are then found in all image pairs by wide-baseline stereo matching and used in a scene structure and camera reconstruction step that can cope with occlusions and outliers [1]. Image pairs suitable for dense matching are automatically selected, rectified and used in dense binocular matching [2]. The dense point cloud obtained as the union of all pairwise reconstructions is fused by a local approximation using oriented geometric primitives [3]. For texturing, every primitive is mapped on the image with the best resolution.


This research was supported by The Czech Academy of Sciences under project 1ET101210406 and by the EU projects eTRIMS FP6-IST-027113 and DIRAC FP6-IST-027787. Richard Szeliski from Microsoft Research provided the ICCV’05 Contest data. Jana Kostková from the Czech Technical University provided routines for dense stereo. Our bundle adjustment routine was based on publicly available software by M.I.A. Lourakis and A.A. Argyros. Frederik Schaffalitzki provided the code for the six-point algorithm. Ondřej Chum provided the code for epipolar geometry estimation unaffected by a dominant plane.


[1] D. Martinec and T. Pajdla. 3D Reconstruction by Gluing Pair-wise Euclidean Reconstructions, or “How to Achieve a Good Reconstruction from Bad Images”. 3DPVT 2006, University of North Carolina, Chapel Hill, USA, June 2006.

[2] J. Kostková and R. Šára. Stratified Dense Matching for Stereopsis in Complex Scenes. In BMVC 2003: Proceedings of the 14th British Machine Vision Conference, volume 1, pages 339-348, Norwich, UK, September 2003.

[3] R. Šára and R. Bajcsy. Fish-Scales: Representing Fuzzy Manifolds. Proc. IEEE Conf. ICCV ’98, pp. 811-817, Bombay, India, January 1998.

[4] H. Cornelius, R. Šára, D. Martinec, T. Pajdla, O. Chum, J. Matas. Towards Complete Free-Form Reconstruction of Complex 3D Scenes from an Unordered Set of Uncalibrated Images. SMVP/ECCV 2004, vol. LNCS 3247, pp. 1-12, Prague, Czech Republic, May 2004. (presentation)

Leave a Reply

You must be logged in to post a comment.