Skeletonization of Petroglyph Shapes
|Place Published||St. Pölten|
|Supervising Tutor||Markus Seidl|
In this thesis, we examine different skeletonization and skeleton pruning algorithms for their representation capabilities of petroglyph shapes. Petroglyphs are forms and figures pecked into rock surfaces by people thousands of years ago. They can be found all over the world and are studied by archeologists with great effort. To support the work of the archeologists on the recognition of shape similarity of petroglyphs, the vision beyond this thesis is to build an automated shape classificator. An important prerequisite for shape classification is the computation of a skeleton. We therefore investigate the origins and the definition of a proper skeleton, and carry out an extensive literature research on skeleton computation algorithms. We compile a test dataset of petroglyph shapes and examine their properties and special needs for the purpose of skeletonization. Subsequently we develop an automated preprocessing pipeline for petroglyph shapes and improve several state-of-the-art skeleton pruning algorithms. At last we evaluate our preprocessing pipeline and the different skeletonization and skeleton pruning algorithms on our full dataset and measure their performance. The proposed preprocessing pipeline produces smooth shapes for nearly 80% of all figures and we observe that a thinning of the shapes constitutes a proper tradeoff between the generation of spurious skeleton branches and the deletion of branches of visually important shape parts for nearly 85% of our shapes.
The work for this thesis has been carried out in course of the project 3D-PITOTI, which is funded from the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement no 600545; 2013-2016.