Geometric Structure of Degeneracy for Multi-body Motion Segmentation

Many techniques have been proposed for segmenting feature point trajectories tracked through a video sequence into independent motions. It has been found, however, that methods that perform very well in simulations perform very poorly for real video sequences. This paper resolves this mystery by analyzing the geometric structure of the degeneracy of the motion model. This leads to a new segmentation algorithm: a multi-stage unsupervised learning scheme first using the degenerate motion model and then using the general 3-D motion model. We demonstrate by simulated and real video experiments that our method is superior to all existing methods in practical situations.

[1]  Kenichi Kanatani,et al.  Geometric Information Criterion for Model Selection , 1998, International Journal of Computer Vision.

[2]  Takeo Kanade,et al.  A Multibody Factorization Method for Independently Moving Objects , 1998, International Journal of Computer Vision.

[3]  C Tomasi,et al.  Shape and motion from image streams: a factorization method. , 1992, Proceedings of the National Academy of Sciences of the United States of America.

[4]  K. Kanatani,et al.  Factorization without Factorization: Complete Recipe , 2004 .

[5]  Y. Wu.  Multibody grouping via orthogonal subspace decomposition, sequences under affine projection , 2001 .

[6]  T. Idehara,et al.  MEMOIRS OF THE FACULTY OF ENGINEERING , 1977 .

[7]  Peter J. Rousseeuw,et al.  Robust regression and outlier detection , 1987 .

[8]  Kohei Inoue,et al.  Separation of Multiple Objects in Motion Images by Clustering , 2001, ICCV.

[9]  Kenichi Kanatani,et al.  Outlier removal for motion tracking by subspace separation , 2002 .

[10]  Peter J. Rousseeuw,et al.  Robust Regression and Outlier Detection , 2005, Wiley Series in Probability and Statistics.

[11]  Takeo Kanade,et al.  A Paraperspective Factorization Method for Shape and Motion Recovery , 1994, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Naoyuki Ichimura.  Motion segmentation based on factorization method and discriminant criterion , 1999, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[13]  C. W. Gear,et al.  Multibody Grouping from Motion Images , 1998, International Journal of Computer Vision.

[14]  Kenichi Kanatani,et al.  Uncertainty modeling and model selection for geometric inference , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[15]  Václav Hlavác,et al.  Ten Lectures on Statistical and Structural Pattern Recognition , 2002, Computational Imaging and Vision.

[16]  Kenichi Kanatani,et al.  Motion segmentation by subspace separation and model selection , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[17]  Kenichi Kanatani,et al.  Evaluation and Selection of Models for Motion Segmentation , 2001, ECCV.

[18]  Naoyuki Ichimura,et al.  Motion segmentation using feature selection and subspace method based on shape space , 2000, Proceedings 15th International Conference on Pattern Recognition. ICPR-2000.

[19]  Kenichi Kanatani,et al.  Motion Segmentation by Subspace Separation: Model Selection and Reliability Evaluation , 2002, Int. J. Image Graph..

[20]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[21]  Kenichi Kanatani,et al.  Automatic Camera Model Selection for Multibody Motion Segmentation , 2002, MVA.

[22]  N. Otsu.  A threshold selection method from gray level histograms , 1979 .