Inference of Multiple Subspaces from High-Dimensional Data and Application to Multibody Grouping

Multibody grouping is a representative of applying subspace constraints in computer vision tasks. Under linear projection models, feature points of multibody reside in multiple subspaces. We formulate the problem of multibody grouping as multiple subspace inference from highdimensional data space. The theoretical value and practical advantage of this formulation come from the relaxation of the motion independency assumption which has to be enforced in most factorization based methods. In the proposed method, an Oriented-Frame (OF), which is a multidimensional coordinate frame, is associated with each data point indicating the point's preferred subspace structure. Then, a similarity measurement of these OFs is introduced and a novel mechanism is devised for conveying the information of the inherent subspace structure among the data points. In contrast to the existing factorization-based algorithms that can not find correct segmentation of correlated motions such as articulated mot...

• Type: Conference
• Year: 2004
• Where: CVPR
• Authors: Zhimin Fan, Jie Zhou, Ying Wu
• Keywords: Computer Vision, CVPR 2004, Inherent Subspace Structure, Multibody Grouping, Multiple Subspace Inference, Subspace Constraints, Subspace Inference Method