W I L L K O M M E N
	Universität Freiburg - Institut für Informatik Lehrstuhl für Mustererkennung und Bildverarbeitung (LMB)

LMB People Research Studien-, Diplom- und Bachelor... () Algorithmen zur Klassifikation... () Trifocal Tensor Estimation Evaluation of Blind Deconvolut... () Matching of crystal surfaces () Lernen von visuellen Objektkla... () Comparison of Polar Fourier Tr... () Subpixelgenaue Bewegungsschaet... () Konturdetektion in zeitaufgelo... () Lab Course I Project: Bildrektifikation () Project: Active Countour und S... () Lectures Events Publications Awards Jobs/Vacancies TILDA Software Interna ()  Printer optimized version of this page [keine Deutsche Seite verfügbar]

ALBERT-LUDWIGS-UNIVERSITÄT FREIBURG INSTITUTE FOR COMPUTER SCIENCE Chair for Image Processing and Pattern Recognition Prof. Dr.-Ing. Hans Burkhardt Georges-Köhler-Allee 52, Room 01-029, D-79110 Freiburg, Tel. 0761-203-8260 The chair of Pattern Recognition and Image Processing is currently offering a Diplomarbeit (Master thesis) on Trifocal Tensor Estimation One of the main objectives of Computer Vision is reconstructing three-dimensional models of scenes and objects from two-dimensional images. 3D reconstruction has applications in fields like robotics, video compression, architectural surveying etc. Making extensive use of Projective Geometry, several techniques have appeared in the last decade that are capable of recovering the complete three-dimensional information from perspective images [2]. An important feature of these algorithms is that no knowledge of the cameras's positions and internal parameters is required: They can use uncalibrated images and produce so called projective reconstructions. For the case of three given images, all projectively relevant camera parameters are encapsulated in a single mathematical object, the trifocal tensor. The tensor is all what is needed for projective reconstruction and can be estimated from image measurements (correspondences) alone. However, reliable estimation of the trifocal tensor is crucial for 3D reconstruction from uncalibrated cameras. Linear estimation is possible but not satisfactory because it does not enforce the nonlinear constraints that must be fulfilled by a valid tensor [1]. The scope of the present master thesis comprises an implementation of the nonlinear method proposed in [3] for trifocal tensor estimation and its comparison with the Gold Standard [2]. Literature: [1] N. Canterakis: A Minimal Set of Constraints for the Trifocal Tensor, in: D. Vernon, editor, European Conference on Computer Vision - ECCV 2000, Springer LNCS 1842, pp. 84-99. [2] R. I. Hartley and A. Zisserman: Multiple View Geometry in Computer Vision, Cambridge University Press, 2000. [3] B. Matei, B. Georgescu and P. Meer: A Versatile Method for Trifocal Tensor Estimation, in: 8th International Conference on Computer Vision, Vancouver, Canada, July 2001, vol. II, pp. 578-585. Candidates: Students in computer science, mathematics or physics. Knowledge of a high-level programming language is desirable. Please contact: N. Canterakis Room: 01-047 Telephone: 0761/203-8269 E-Mail: canterakis@informatik.uni-freiburg.de Oct. 2006