Visual Computing for Archaeological Artifacts with Integral Invariant Filters in 3D

Abstract

3D-artifacts from ancient civilizations contain many different kinds of information in form of forensic trace evidence, e.g., tool marks from styli or fingerprints on wax sealings. These very fine structures are increasingly captured by various 3D-acquisiton techniques and stored as irregular meshes. We introduce filter algorithms for the processing of these datasets to finally extract meaningful information at predefined scales. Therefore, Multiscale-Integral Invariants (MSII) are introduced as robust filter methods with their four different variants, using volume, patch, surface and line integrals for their specific sensitivity on mean curvature, Gaussian curvature or noise detection. Smoothing as known from 2D-raster image processing cannot be applied directly. It needs adaptation to the irregular structure of the triangular grids describing 2D-manifolds in 3D-space.We introduce a fast 1-ring smoothing with a skillful weighting by distance and area of the neighboring points and triangles. Finally, we apply our technique to the various motivating examples for showing the results as false color images with isolines, indicating the respective field of function values, e.g., curvature in various norms or correlations in the feature space. Smooth isolines are indicators for the successful removal of noise. We finally compare the fully automated results with a manual graphic rendering of a faded handwriting found in the tomb of the empress Gisela of Swabia.