Semi-automated Annotation of Repetitive Ornaments on 3D Painted Pottery Surfaces

Abstract

The creation of drawings from the surface of painted pottery artifacts is an important practice in archaeological research and documentation. Traditional approaches include manual drawings using pen and paper, either directly on the physical surface, or from photographs, while more recent approaches are supported by photography or flattening of 3D digitized objects. Elaborate vase paintings, mostly showing figural scenes, often comprise ornamental decorations in secondary position or in the background, exhibiting repetitive patterns. We propose a tool supporting the creation of archaeological drawings with a semi-automatic extraction of ornamental surface sections, based on a combination of user-defined queries and self-similarity detection. Appropriate heuristics allow to detect the presence and positions of ornamental bands, a frequently occurring scheme, where ornamental primitives are evenly spaced along the tangential direction of a vessel's solid of revolution. Our interactive tool allows domain experts to efficiently select ornamental queries, and assess the quality of resulting similarity detections. First experiments with real world artifacts from ancient Greek and Peruvian cultures confirm the feasibility of the approach.