Estimation of Similarity Based on Comparing partial Shapes of 3D Objects

Comparing object shapes is essential for many computer vision tasks such as object model categorization and object recognition. This paper presents a method for comparing an object reconstructed by faces viewing from a specified point with a geometric model. The basic idea of the method is as follows: (1) The convex hull of the viewed faces is constructed in order to supplement the part which is not viewed from the specified point. (2) The viewed faces are compared with faces of the geometric model, and the convex hull of the faces corresponding to the viewed faces is made. (3) The numerical similarity among these two convex objects is calculated according to the algorithm for determining similarities of convex polyhedra described in the previous report. Various cases of object shapes have been tested. The experimental results show that this method is good for recognizing shape similarity based on viewing faces.