deformable graph matching - strongfstrong-zhou.com.pdfVIP

Deformable Graph Matching Feng Zhou Fernando De la Torre Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213 ftorre@ Abstract Graph matching (GM) is a fundamental problem in com- puter science, and it has been successfully applied to many problems in computer vision. Although widely used, exist- ing GM algorithms cannot incorporate global consistence among nodes, which is a natural constraint in computer vision problems. This paper proposes deformable graph matching (DGM), an extension of GM for matching graphs subject to global rigid and non-rigid geometric constraints. The key idea of this work is a new factorization of the pair-wise affinity matrix. This factorization decouples the affinity matrix into the local structure of each graph and the pair-wise affinity edges. Besides the ability to incor- porate global geometric transformations, this factorization Figure 1. Matching two human poses with and features using offers three more benefits. First, there is no need to com- DGM. DGM simultaneously estimates the correspondence and a pute the costly (in space and time) pair-wise affinity ma- smooth non-rigid transformation between nodes. DGM is able to trix. Second, it provides a unified view of many GM meth- factorize the pair-wise affinity matrix as a Kronecker prod- ods and extends the standard iterative closest point algo- uct of six smaller matrices. The first two groups of matrices of size rithm. Third, it allows to use the path-following optimiza- and encode the structure of each of the graphs (i.e., tion algorithm that leads to improved optimization strate- adjacency