Content Based Image Retrieval using Interest Point Algorithms in Context of Scientific Cultural Image Collections of Hebraic Tombstones

The Digital Research Infrastructure for the Arts and Humanities-project (Dariah-DE) is dedicated to evaluate information retrieval technologies for research infrastructures of social, humanand cultural studies like universities. One on of the main project-participants is the Salomon-Ludwig-Steinheim Institute of German-Jewish-History which documents Hebraic tombstones as a part of Jewish history and life. A query-by-example could help to improve investigations in this image-database. The content based image retrieval (CBIR) could be done using different features like interest point algorithms (IPA). These algorithms find the most stable points like corner in images and calculate a comparable representation for this point using the surrounding pixel intensities. An amount of these stable keypoints will describe the content of the image. In this paper an example collection of Hebraic tombstone is used to evaluate IPA-detectordescriptor-pairs like SIFT-SIFT, SIFT-BRISK, SURF-SURF, SURF-SIFT, SURF-BRISK and CenSurE-SIFT. Their tolerances in the difference of object-scale, illumination and perspective angle are tested. Further user-driven test-scenarios for CBIR are used to investigate the applicability of the IPAs when similar images in context of scientific cultural research have to be retrieved.