A Scale Normalization Algorithm Based on MR-GDS for Archaeological Fragments Reassembly

Congli Yin, Pengbo Zhou, Mingquan Zhou, Zhongke Wu, Guoguang Du

Abstract: In template-based archaeological fragments reassembly, the fragments and the template usually have scale variances, which needs to be eliminated by scale normalization. Traditional methods need at least two pairs of matched points for scale normalization, which is nontrivial for many fragments. In this paper, a scale normalization algorithm for template-based fragments reassembly is proposed, which solves the reassembly problem for fragments and the template with scale variances. Firstly, one pair of corresponding points between the fragment and the template is selected as the anchor points. Secondly, the Geodesic Disk Spectrum (GDS) of the fragment and the Multi-Radii Geodesic Disk Spectrum (MR-GDS) of the template are computed, respectively. Thirdly, the GDS descriptors are compared to obtain candidates of the scale variances, which will be further verified through an ICP-based strategy. At last, the fragments are registered to the template after scale normalization. Experiments on scale normalization and fragments reassembly all achieved accurate results. The proposed algorithm is suitable for the scale estimation of small fragments, which expands the application scenarios greatly for the template-based fragments reassembly.