Plane-curve-based matching for broken bronze mirror reassembling

Abstract

Bronze mirror is one of the most famous bronze artifacts in ancient China, some of which have been broken into several small fractures after excavation. Recently, computer scientists collaborating with archeologists focused on fractures matching automatically according to geometry curve. In this paper, a novel method is proposed for several fractures automatic reassembling to improve the speed and accuracy. Firstly, the one-shot image is utilized to collect image data for broken bronze mirror by digital camera. Secondly, watershed algorithm is used to segment and mark each fracture. Thirdly, the longest common curve is found by combining corners detection, coarse matching and fine matching, taking length, angle and curvature into account. The curvature consistency is used to omit the outlier mirror external curve to guarantee matching correctly by circle-shaped structure. Finally, the least square method is performed to compute rigid transformation to reassemble neighbor fractures by matching increment. Experimental results on broken bronze mirror demonstrate the correctness and robustness of our method.