Tomas Pajdla   [Pie-dlah]

Assistant Professor

 

Address:

Karlovo namesti 13, 121-35 Praha 2, 
Czech Republic
Tel.: +420-224-357-348
Fax: +420-224-357-385
pajdla\cmp_felk-cvut=cz

Center for Machine Perception
Department of Cybernetics
Faculty of Electrical Engineering
Czech Technical University in Prague
Research: Geometry of  Non-central, Omnidirectional, and Panoramic Vision.

NON-CENTRAL

OMNI

 PANORAMIC

MULTI-VIEW

 IBR
Teaching: Computer Vision & Intelligent Robotics
"How to" for my students,  Master, Batchelor, and semestral projects (in Czech)
Projects: DIRAC Detection & Identification of Rare Audio-Visual Cues, FP6 EU FP6-IST-027787 
BeNoGo Being There - Without Going, FP5 EU IST-2001-39184
OMNIVIEWS Omni-directional Visual System, FP5 EU IST-1999-29017
Lectures:

Omnidirectional Vision Course @ ICCV 2003  
(epipolar geometry estimation of  dioptric & catadioptric omnidirectional cameras, stereo geometry of non-central cameras)
Stereo Geometries of Non-central Cameras @ Computer Vision Colloquium 2001   
(stereo geometry of non-central cameras, nice linear cameras generated by collineations in P³)

Students: Finished: T.Svoboda (PhD), M.Urban (PhD), S.Gaechter (MSc), O.Chum (MSc), R.Horcik (MSc), J.Sivic (MSc), M.Menem (MSc), B.Micusik (PhD). Current: H.Bakstein (PhD),  D.Martinec (PhD), M.Havlena (PhD), T.Ehlgen (PhD), Z.Kukelova (PhD).
Service: PC Chair of ECCV 2004, Area Chair of  ICCV 2005, BMVC 2005, ACCV 2006, CVPR 2006, BMVC 2006, PC Member of OMNIVIS workshops, Reviewer of IEEE PAMI & IJCV
Links: Short Curriculum Vitae, Long Curriculum Vitae, CMP spin-off  Neovision Ltd.
Publications:  
Non-Central T.Pajdla. Non-classical Ray Cameras. Technical report CTU-CMP-1999-11. November 1999. (programme to study non-central cameras - mosaics, plenoptic functions, light fields - formulated)
F.Huang, T.Pajdla. Epipolar Geometry in Concentric Panoramas. Technical report CTU-CMP-2000-07, March 2000 (stereo geometry of one configuration of concentric panoramas - inspiring but not entirely correct)
T.Pajdla. Epipolar Geometry of Some Non-classical Cameras. Computer Vision Winter Workshop 2001. Bled, Slovenia. pp. 223--233, February 2001. (generalized epipolar geometry implies that rays of both cameras must form opposite reguli)
 H.Bakstein, T.Pajdla. 3D Reconstruction from 360x360 Mosaics. CVPR 2001, pp. 72-77, IEEE December 2001. (classes of reconstruction from uncalibrated 360x360 mosaic, epipolar alignment)
 T.Pajdla. Stereo with Oblique Cameras. Workshop on Stereo and Multi-Baseline Vision, pp. 85-91, IEEE December 2001. (see the following IJCV paper)
 J.Sivic, T.Pajdla. Geometry of Concentric Multiperspective Panoramas. Technical report CTU-CMP-2002-05. February 2002. (geometry and stereo-geometry, classification of generalized epipolar geometries)
 T.Pajdla. Geometry of Two-Slit Camera. Technical report CTU-CMP-2002-02, March 2002.  (geometry of X-slits
cameras and Oblique cameras, analyzed in the complexification of P^3)
 T.Pajdla. Stereo with Oblique Cameras. IJCV, 47(1):161-170, Kluwer May 2002. (the most non-central camera with a generalized epipolar geometry: definitions, properties, philosophy, genaration by a collineation)
H.Bakstein, T.Pajdla. Rendering Novel Views from a Set of Omnidirectional Mosaic Images. Workshop on
Omnidirectional Vision and Camera Networks 2003, CD ROM, IEEE June 2003.  (image beased rendering with non-central cameras)
 H.Bakstein, T.Pajdla, D.Vecerka. Rendering Almost Perspective Views from a Sparse Set of Omnidirectional Images. BMVC 2003, pp. 241--250, BMVA September 2003. (IBR with non-central images - X-Slits images can be scaled to look like perspective ones)
D.Feldman, T. Pajdla, D.Weinshall. On the Epipolar Geometry of the Crossed-Slits Projection. ICCV 2003, pp. 988--995, IEEE October 2003. (stereo geometry of  X-Slits cameras, search curves, fundamental matrix)
 B.Micusik, T.Pajdla. Autocalibration & 3D Reconstruction with Non-central Catadioptric Cameras . CVPR 2004, Washington US, June 2004. (autocalibration of non-central quadric catadioptric cameras by epipolar geometry estimation, worked out for parbolic, hyperbolic, and spherical mirrors.)
 M.Menem, T.Pajdla.Constraints on Perspective Images and Circular Panoramas. BMVC 2004, BMVA, September 2004.  (formutalion and estimation of the multiview constrint for the mixed camera pair, surprisingly simple)

Omni-directional 
Panoramic

 T.Svoboda, T.Pajdla, V.Hlavac. Epipolar Geometry for Panoramic Cameras. ECCV 1998, Springer LNCS 1406, pp. 335-340, June 1998. (the first formulae for a general epipolar geometry for central camtadioptric cameras ... hyperbolic mirror only)
 T. Svoboda, T.Pajdla, V.Hlavac. Motion Estimation Using Central Panoramic Cameras. IEEE Int. Conf. on Intelligent Vehicles 1998, pp. 335-340, Causual Productions October 1998. (cameras with wide field of view provide very reliable camera motion from epipolar geometry)
 T.Pajdla, T.Svoboda, V.Hlavac. Epipolar Geometry of Central Panoramic Cameras. In Panoramic Vision: Sensors, Theory, and Applications, pp. 85-114. Springer Verlag, 2001. (Overview of many central catadioptric cameras, definition of panoramic & omnidirectional camera, epipolar geometry for hyperbolic and parabolic mirrors, point normalization for epipolar geometry estimation with omni-cameras)
 S-K.Wei, M.Urban, T.Pajdla. Stereo Matching of Catadioptric Panoramic Images. Technical report CTU-CMP-2000-08. March 2000. (epipolar alignment for central catadioptric images, search for correspondences by dynamic programming ... complete-field of view leadds to better results)
S. Gaechter, T. Pajdla, B.Micusik. Mirror Design for an Omnidirectional Camera with a Space Variant Imager. Workshop on Omnidirectional  Vision 2001(ICAR 2001), pp. 99-105, IEEE August 2001. (uniform resolution catadioptric sensor with SVAVISCA imager)
T. Svoboda, T.Pajdla. Epipolar Geometry for Central Catadioptric Cameras. IJCV, 49(1):23-37, Kluwer August 2002. (parabolic+hyperbolic+elliptic nirrors = all cases done)
 H.Bakstein, T.Pajdla. Panoramic Mosaicing with a 180 deg Field of View Lens. Workshop on Omnidirectional  Vision 2002, pp. 60--67, IEEE June 2002. (model of Nikon FC-E8 lens, 360x360 photograpnic quality mosaic)
 G.Sandini, J.Santos-Victor, T.Pajdla, F. Berton. OMNIVIEWS: Direct Omnidirectional Imaging Based on a Retina-line Sensor. IEEE International Conference on Sensors 2002, IEEE 2002. (SVAVISA retina-like imager combined with a suitably designed mirror - optimal resolution for low bandwidth)
B.Micusik, T.Pajdla. Estimation of Omnidirectional Camera Model from Epipolar Geometry. CVPR 2003, pp. 485-490, IEEE June 2003. (auto-calibration of very wide-angle-of-view (i.e. omnidirectional) lens  by epipolar geometry estimation, 9 point RANSAC by Polynomial eigenvalue problem from a oinearization... for Nikon FC-E8 lens)
B.Micusik, T.Pajdla. Omnidirectional Camera Model and Epipolar Geometry Estimation by RANSAC with Bucketing. SCIA 2003, pp. 83-90, IEEE June 2003. (RANSAC improvement bu BUCKETING in the above method)
B.Micusik, T.Pajdla. Para-catadioptric Camera Auto-calibration from Epipolar Geometry. ACCV 2004, Korea January 2004. (autocalibration of a para-catadioptric camera by epipolar geometry estimation can be done as a Quartic polynomial eigenvalue problem, no linearization neecessary, done for hyperbolic mirror as well)
B.Micusik, D.Martinec, T.Pajdla. 3D Metric Reconstruction from Uncalibrated Omnidirectional Images. ACCV 2004, Korea January 2004. (automatic correspondece, autocalibration, and metric reconstruction from many wide base-line wide-angle-of-view (Canon EOS-D1 with Sigma 180x360 lens) images)
H.Bakstein, T.Pajdla. Calibration procedure for a 360x360 mosaic camera. International archives of photogrammetry, remote sensing and spatial information sciences, XXXVI-5/W8:online, February 2005.
B.Micusik, T.Pajdla. Structure from Motion with Wide Circular Field of View Cameras. IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 7, pp. 1135-1149, Jul, 2006.

Factorization
Many-View 
Reconstruction

 

 D.Martinec, T.Pajdla. Structure from Many Perspective Images with Occlusions. ECCV 2002, pp. 355-369, Springer LNCS, May 2002. (projective factorization with missing data = projective depth estimation using Sturm & Triggs method + Extension of Jacobs filling method from affine to perspective camera)
D.Martinec, T.Pajdla. Line Reconstruction from Many Perspective Images by Factorization. CVPR 2003, pp. 497-502, IEEE June 2003. (projective factorization with missing data & outliers ... a little of RANSAC added to the above method)
D.Martinec, T.Pajdla. Consistent Multi-View Reconstruction from Epipolar Geometries with Outliers. SCIA 2003, pp. 493-500, IEEE June 2003. (projective factorization can be done also for lines using Plucker coordinates)
 D.Martinec, T.Pajdla. 3D Reconstruction by Fitting Low-Rank Matrices with Missing Data. CVPR 2005, pp. 198-205, IEEE June 2005. (linear formulation for computing the logarithms of projective depths,  linear formulation for consistent reconstruction)
T.Svoboda, D.Martinec, and T.Pajdla. A convenient multi-camera self-calibration for virtual environments.  PRESENCE: Teleoperators and Virtual Environments, 14(4):407-422, August 2005.
Orientation

T. Werner, T.Pajdla, V. Hlavac. Oriented Projective Reconstruction. ÖAGM/IAPR, pp. 245-254, May 1998. Best paper award. (Hartley's cheriality rediscovered, later but independently, omnidirectional cameras have hapl-rays, and thus are oriented!)
 T.Werner, T.Pajdla. Cheirality in Epipolar Geometry. ICCV 2001, pp. 548-553, IEEE July 2001. (A stronger Epipolar Constraint formulated for cameras with half-rays) 
 T.Werner, T.Pajdla. Oriented Matching Constraints. BMVC 2001, pp. 441-450, BMVA September 2001.(Omnidirectional camera centers are inside the convex hull of reconstructed points, more on orientation constraint for lines, tensorial language). 
 O.Chum, T.Werner, T.Pajdla. Joint Orientation of Epipoles. BMVC 2003, pp. 73-82, BMVA 2003. (epipolar half-lines may help to disambiguate image matching, horpoter is only a curve segment)

Correspondence
Localization
Rendering
Range		 O.Chum, T.Pajdla, P.Sturm. The Geometric Error for Homographies. Computer Vision and Image Understanding Volume 97, Issue 1 , January 2005, Pages 86-102. 
 J.Matas, O.Chum, M.Urban, T.Pajdla. Robust Wide Baseline Stereo from Maximally Stable Extremal Regions.Image and Vision Computing, 22(10):761-767, September 2004. (The journal version of BMVC 2002 paper.)
 J.Matas, O.Chum, M.Urban, T.Pajdla. Robust Wide Baseline Stereo from Maximally Stable Extremal Regions. BMVC 2002, pp. 384-393, BMVA Septmeber 2002. (formulation of a robust matching paradigm: Distinguished vs. Measurement regions + ordering-based similarity & RANSAC-based estimation of multiview geometry) Best paper award.
 M.Urban, T.Pajdla, V.Hlavac. Projective Reconstruction From N Views Having One View in Common. Vision Algorithms, LNCS 1883, pp. 116-131, September 1999. (projective reconstruction from many views with one view in common ... cake configuration) 
T.Werner, T.Pajdla, V.Hlavac. Efficient 3-D Scene Visualization by Image Extrapolation. ECCV 1998, Springer LNCS 1406, pp. 382--395, June 1998. (novel view synthesisby image interpolation is simpler than by image estrapolation, which is equivalent to reconstructing a the scene)
 T.Werner, T.Pajdla, V.Hlavac, A.Leonardis, M.Matousek. Selection of Reference Images for Image-Based Representation. Computing 68(2):163--180, Springer March 2002. (reference views can be selected using reprojection error)
 T. Pajdla, V.Hlavac. Zero Phase Representation of Panoramic Images for Image Based Localization. CAIP 1999, Springer LNCS 1689, pp. 550-557. September 1999. (image based compas from phase (Zero Phase Representation) of periodic panoramic images used for robot localization)
 T.Pajdla. Camera Calibration and Euclidean Reconstruction from Known Observer Translations. CVPR 1998, pp. 421-426, IEEE June 1998. (linear ethod for camera-on-robot calibration using controlled translcations)
 P.Krsek, T.Pajdla, V.Hlavac. Differential Invariants as the Base of Triangulated Surface Registration. Computer Vision and Image Understanding 87(1-3):27-38, Academic Press, July 2002. (differential invariants of triangulated surfaces)
T.Pajdla, L.Van Gool. Matching of 3-D Curves Using Semi-differential Invariants. ICCV 1995, pp. 390--395, IEEE June 1995. (curve matching with ICRP = Iterative Closest Reciprocal Point Algorithm)
T.Pajdla, V.Hlavac, R.Sara. Segmentation of Range Images. Acta Stereologica, 13(2):459-464, June 1994.