Grouping visual objects based on flow structure
-
1
Goethe University, BFNT Frankfurt, Germany
In pursuit of a fast motion estimation scheme for the simultaneous estimation of motion and segmentation in moving visual scenes, we develop a two-step approach that first estimates a sparse set of reliable motion vectors at carefully selected locations, and then clusters these vectors according to consistent flow models. Such models in turn determine dense parametric flow fields from the grouped sparse motion vectors. For the first task, namely, the sparse flow field estimation, an enhanced local motion estimator [1] is used which, in addition to local motion estimates, it provides accuracy figures in terms of individual covariance matrices. The clustering procedure can be considered as an enhanced iterative optimization scheme in the EM-spirit, incorporating both prior knowledge on the flow fields (currently: second order polynomials with an additive curvature constraint), and the covariance matrices for the local flow measurements. In contrast to other recent grouping methods such as [2], our method is much simpler and computationally cheaper. The processing scheme is efficient in that measurements are made only on prominent points, grouping is performed on such sparse, yet informative data, and full-resolution can be obtained by associating each pixel to one of some few motion hypotheses. Finally, the approach is tested on image sequences taken from the Middlebury standard database (see Fig. 1).
Figure 1. (Left) The original “Venus” input image. (Right) Color-coded clustering result obtained with the proposed algorithm. The red blocks indicate the spatial center of each cluster.
References
1. Mester R and Hötter M. Robust displacement vector estimation including a statistical error analysis. In Int. Conf. on Image Processing and its Applications, 168-172 (1995).
2. Ren X. Local grouping for optical flow. In CVPR, 1-8 (2008).
Keywords:
computational neuroscience
Conference:
Bernstein Conference on Computational Neuroscience, Berlin, Germany, 27 Sep - 1 Oct, 2010.
Presentation Type:
Presentation
Topic:
Bernstein Conference on Computational Neuroscience
Citation:
Guevara
A,
Conrad
C and
Mester
R
(2010). Grouping visual objects based on flow structure.
Front. Comput. Neurosci.
Conference Abstract:
Bernstein Conference on Computational Neuroscience.
doi: 10.3389/conf.fncom.2010.51.00003
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
02 Sep 2010;
Published Online:
22 Sep 2010.
*
Correspondence:
Dr. Alvaro Guevara, Goethe University, BFNT Frankfurt, Frankfurt am Main, Germany, alvaro.guevara@tu-dresden.de