Detection of Imbalanced Vertices in 3D Meshes

A critical task in corner detection in 2D images is on the distinction between a corner pixel and a pixel with a large gradient (i.e., an edge pixel). Imbalanced point detection was proposed to address this problem, where a corner pixel is characterized as a pixel with an imbalanced appearance, while an edge pixel has the opposite property. With extensive experiments, an imbalanced point detector has been shown to outperform many state-of-the-art detectors in terms of high repeatabilities and localization accuracy. A critical task in corner detection in 3D meshes is on the distinction between a corner vertex and a vertex with a large curvature. In this paper, we generalize the idea of imbalance oriented corner detection from 2D images to 3D meshes. We propose four algorithms on the detection of imbalanced vertices in 3D meshes. We present an experimental study of the four proposed algorithms and two curvature-based methods. Results show the promise of the proposed method on detection of 3D imbalanced vertices.