Ridge Point Extraction with Non-Maximum Suppression on Irregular Grids
Richard Schönpflug, Hubert Mara · heiDOK (Heidelberg University) · 2016
Assyriology is the study of cultures related to cuneiform writing, which was used for more than three millennia before Christ in the ancient Middle East.Drawing hundreds of thousands of documents with cuneiform script manually is a tedious task and leads to a demand for automated tools assisting the daily work of assyriologists.The cuneiform script is a handwriting using wedges (Latin: cunei) imprinted into clay tablets.Therefore the digitization of cuneiform tablets is increasingly using 3Dscanners that provide irregular triangular grids in R 3 .These grids i.e. meshes are discrete manifolds, which are first filtered by using Multi-Scale Integral Invariants (MSIIs) for visualization.Secondly the MSII filter results are used to extract points along the or ridges within the 3D-model leading to a digital drawing of e.g. a cuneiform tablet.Therefore we choose the idea of the non-maximum suppression as used by the Canny edge detector for raster images.In contrast to the Canny edge detector we had to (i) to adapt to an arbitrary number of neighboring vertices, which have to be reduced locally in case of flat areas; (ii) to implement an estimator for the gradient direction, which cannot be provided by the MSII filter; and (iii) to provide a border treatment as real world meshes have missing parts.All the work was embedded within our modular GigaMesh software framework.Results are shown for synthetic and real data, demonstrating a computational complexity of O(n), which requires only one parameter.Finally a summary and an outlook are given.