Vectorization of 3D-Characters by Integral Invariant Filtering of High-Resolution Triangular Meshes

Abstract

Motivated by the demand of today's Assyriologists we develop a system for automated detection and extraction of cuneiform script, which is one of the most important sources for ancient history. Traditional means of documentation are (i) photographs and (ii) manual drawings, which are increasingly replaced by shape acquisition using 3D-scanners resulting in (iii) high-resolution 3D-models. To utilize the full potential of the acquired 3D-data, we propose a filtering algorithm on 2D-manifolds using Multi-Scale Integral Invariants (MSII) to detect characters within a high-dimensional feature space. As MSII filtering is a local method it overcomes the drawbacks of global illumination methods using virtual light sources. This filtering technique allows for rendering false-color images of the tablets without shadowing effects making the tablets already easy to read. With an additional step of the processing pipeline of our software framework \emph{GigaMesh}, we can extract vector drawings. These are the basis for character recognition as well as for future paleographic analysis. The vectorized characters are stored in the XML-based \emph{Scalable Vector Graphics} (SVG) format. This results in a tremendous reduction of the triangular mesh data to a meaningful spline representation of the tablets' contents.