GPCA with denoising: A moments-based convex approach

This paper addresses the problem of segmenting a combination of linear subspaces and quadratic surfaces from sample data points corrupted by (not necessarily small) noise. Our main result shows that this problem can be reduced to minimizing the rank of a matrix whose entries are affine in the optimization variables, subject to a convex constraint imposing that these variables are the moments of an (unknown) probability distribution function with finite support. Exploiting the linear matrix inequality based characterization of the moments problem and appealing to well known convex relaxations of rank leads to an overall semi-definite optimization problem. We apply our method to problems such as simultaneous 2D motion segmentation and motion segmentation from two perspective views and illustrate that our formulation substantially reduces the noise sensitivity of existing approaches.

[1]  David Suter,et al.  Two-View Multibody Structure-and-Motion with Outliers through Model Selection , 2006, IEEE Trans. Pattern Anal. Mach. Intell..

[2]  John B. Moore,et al.  A Newton-like method for solving rank constrained linear matrix inequalities , 2006, 2004 43rd IEEE Conference on Decision and Control (CDC) (IEEE Cat. No.04CH37601).

[3]  Allen Y. Yang,et al.  Estimation of Subspace Arrangements with Applications in Modeling and Segmenting Mixed Data , 2008, SIAM Rev..

[4]  S. Shankar Sastry,et al.  Generalized principal component analysis (GPCA) , 2005, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[5]  P. Anandan,et al.  Integrating Local Affine into Global Projective Images in the Joint Image Space , 2000, ECCV.

[6]  S. Shankar Sastry,et al.  Two-View Multibody Structure from Motion , 2005, International Journal of Computer Vision.

[7]  J. L McGregor.  Solvability criteria for certain N-dimensional moment problems☆ , 1980 .

[8]  Emmanuel J. Candès,et al.  A Singular Value Thresholding Algorithm for Matrix Completion , 2008, SIAM J. Optim..

[9]  Peter Meer,et al.  Simultaneous multiple 3D motion estimation via mode finding on Lie groups , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.

[10]  J. Lasserre,et al.  Detecting global optimality and extracting solutions in GloptiPoly , 2003 .

[11]  René Vidal,et al.  A Benchmark for the Comparison of 3-D Motion Segmentation Algorithms , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.

[12]  René Vidal,et al.  Combined central and subspace clustering for computer vision applications , 2006, ICML.

[13]  Raul E. Curto,et al.  Truncated K-moment problems in several variables , 2005 .

[14]  Stephen P. Boyd,et al.  Enhancing Sparsity by Reweighted ℓ1 Minimization , 2007, 0711.1612.

[15]  Stephen P. Boyd,et al.  Log-det heuristic for matrix rank minimization with applications to Hankel and Euclidean distance matrices , 2003, Proceedings of the 2003 American Control Conference, 2003..

[16]  Jean B. Lasserre,et al.  Global Optimization with Polynomials and the Problem of Moments , 2000, SIAM J. Optim..

[17]  David Suter,et al.  A Model-Selection Framework for Multibody Structure-and-Motion of Image Sequences , 2007, International Journal of Computer Vision.

[18]  David J. Kriegman,et al.  Clustering appearances of objects under varying illumination conditions , 2003, 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings..

[19]  A. Garulli,et al.  Positive Polynomials in Control , 2005 .

[20]  Allen Y. Yang,et al.  Segmentation of hybrid motions via hybrid quadratic surface analysis , 2005, Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1.