3D pottery shape matching using depth map images

Abstract

The utilisation of 3D computer graphics technologies in the domain of pottery analysis can enhance archaeological research in terms of data management, indexing and shape matching. In this paper, we attempt to reduce the dimensions of the 3D vessel shape matching problem in order to create Web-enabled compact shape descriptors applicable for content-based retrieval systems. This is achieved by exploiting significant morphological features of vessels such as the rotational symmetry and the opposed positioning of appendages. We propose the idea of capturing the surface relief of a normalised, in terms of scale, position and orientation 3D vessel on a pair of depthmap images. We question the ability of performing shape matching of complete or nearly complete 3D vessels by encoding depthmap images using 2D shape descriptors. We have evaluated the performance of two shape descriptors applied on depthmap images in relation to a generic 3D shape descriptor. For the performance evaluation, we have implemented multiple queries-by-example on five ancient Greek vessel shape categories found in our calibrated 3D vessel database.