Stratified Generalized Procrustes Analysis

This paper deals with generalized procrustes analysis. This is the problem of registering a set of shape data by finding a reference shape and global rigid transformations given point correspondences. The transformed shape data must align with the reference shape as best possible. This is a difficult problem. The classical approach computes alternatively the reference shape, usually as the average of the transformed shapes, and each transformation in turn. We propose a stratified approach inspired by recent results obtained in Structurefrom-Motion. Our stratified approach offers a statistically grounded framework for obtaining both the transformations and the reference shape at once in two steps. First, we compute a reference shape and affine transformations. Second, we upgrade these transformations to the sought after similarity or euclidean transformations. In practice each of these two steps involves solving a non-convex optimization problem. We provide convex approximations and cl...

• Type: Journal
• Year: 2010
• Where: BMVC
• Authors: Adrien Bartoli, Daniel Pizarro, Marco Loog
• Keywords: BMVC 2010, Computer Vision, Reference Shape, Shape Data, Stratified Approach