A point cloud simplification method using clustering and saliency for cultural heritage reconstruction

Abstract

With the rapid development of 3D scanning technologies, high-density point clouds of cultural heritage artifacts such as stone carvings, statues pose significant challenges in storage, processing, and accurate reconstruction. This paper proposes a point cloud simplification method tailored for cultural heritage applications, combining clustering and saliency analysis to preserve intricate surface details critical for archaeological studies. By segmenting point clouds into clusters with normal vector constraints and evaluating saliency through roughness and curvature metrics, our method adaptively retains primary features including engraved patterns weathered textures while simplifying non-feature regions. Experiments on stone carvings from the Northern Song Imperial Mausoleum, Terracotta Warriors, and Stanford datasets demonstrate that the algorithm effectively avoids mesh holes and maintains geometric fidelity, enabling efficient 3D reconstruction for heritage conservation. This work bridges advanced point cloud processing with practical archaeological needs, offering a robust tool for digitizing and analyzing cultural relics with minimal loss of historically significant details.