Abstract
A ‘Gaussian-blob model’ for local spatiotemporal gray value variations is exploited in order to simultaneously estimate an optical flow vector and the dominant spatiotemporal scale of the gray value variation which forms the basis for this estimation. A two-step, non-iterative, local estimation approach is developed which does not include any explicit smoothness term. Various experiments with real world image sequences nevertheless yield smooth optical flow fields with surprisingly good results, even close to discontinuities of the optical flow field along occluding contours. ‘Close’ in this context means within one to three pixels, i. e. much less than the width of the weight masks employed to estimate the spatiotemporal gradient of the gray value variation.
Preview
Unable to display preview. Download preview PDF.
References
J.L. Barron, D.J. Fleet, and S.S. Beauchemin: Performance of Optical Flow Techniques. International Journal of Computer Vision 12 (1994) 43–77
A. Gehrke: Untersuchungen zur Segmentierung von optischen Flußfeldern. Diplomarbeit, Institut für Algorithmen und Kognitive Systeme, Fakultät für Informatik der Universität Karlsruhe (TH), August 1995 (in German)
B.K.P. Horn: Robot Vision. The MIT Press, Cambridge/MA and London/UK 1986
B. Jähne: Analytical Studies of Low-Level Motion Estimators in Space-Time Images Using a Unified Filter Concept. Proc. IEEE Conference on Computer Vision and Pattern Recognition CVPR '94, 21–23 June 1994, Seattle/WA, pp. 229–236
J.J. Koenderink, A. Kappers, and A. van Doorn: Local Operations: The Embodiment of Geometry. In: G.A. Orban and H.-H. Nagel (Eds.), Artificial and Biological Vision Systems, Springer-Verlag Berlin Heidelberg New York/NY 1992, pp. 1–23
H. Kollnig: Ermittlung von Verkehrsgeschehen durch Bildfolgenauswertung. Dissertation, Fakultät für Informatik der Universität Karlsruhe (TH), Karlsruhe / Germany, February 1995. In: Dissertationen zur Künstlichen Intelligenz DISKI 88, infix, Sankt Augustin / Germany 1995 (in German)
T. Lindeberg: Direct Estimation of Affine Image Deformations Using Visual Front-End Operations with Automatic Scale Selection. Proc. International Conference on Computer Vision ICCV '95, 20–23 June 1995, Cambridge/MA, pp. 134–141
H.-H. Nagel: Analysis of Gradient Outer Product Tensor Fields. Internal Report (Draft, 1995)
M. Otte: Extraktion von linienförmigen Merkmalen und Ermittlung des optischen Flusses mit seinen Ableitungen aus Voll-und Halbbildfolgen. Dissertation, Fakultät für Informatik der Universität Karlsruhe (TH), Karlsruhe/Germany, February 1994. In: Dissertationen zur Künstlichen Intelligenz DISKI 63, infix, Sankt Augustin / Germany 1994 (in German)
M. Otte and H.-H. Nagel: Optical Flow Estimation: Advances and Comparisons. Proc. Third European Conference on Computer Vision ECCV '94, Vol. I, Stockholm / Sweden, 2–6 May 1994, J.-O. Eklundh (Ed.), Lecture Notes in Computer Science 800, Springer-Verlag Berlin Heidelberg New York/NY 1994, pp. 51–60. A revised and expanded version of this contribution will appear in Artificial Intelligence Journal (1995, in press).
C.-F. Westin: A Tensor Framework for Multidimensional Signal Processing. Linköping Studies in Science and Technology, Dissertations No. 348; Department of Electrical Engineering, Linköping University, S-58183 Linköping/Sweden (1994)
J. Zhang and H.-H. Nagel: Texture-Based Segmentation of Road Images. Proc. Intelligent Vehicles '94 Symposium, 24–26 October 1994, Paris/France, pp. 260–265
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1996 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Nagel, H.H., Gehrke, A., Haag, M., Otte, M. (1996). Space- and time-variant estimation approaches and the segmentation of the resulting optical flow fields. In: Li, S.Z., Mital, D.P., Teoh, E.K., Wang, H. (eds) Recent Developments in Computer Vision. ACCV 1995. Lecture Notes in Computer Science, vol 1035. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-60793-5_64
Download citation
DOI: https://doi.org/10.1007/3-540-60793-5_64
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-60793-9
Online ISBN: 978-3-540-49448-5
eBook Packages: Springer Book Archive