Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

International database of facial images for performance and ISO/IEC 19794-5 conformance tests

: Ebinger, Peter; Castro Neves, Margarida; Salamon, René; Seibert, Helmut

Brömme, Arslan (Ed.); Busch, Christoph (Ed.); Hühnlein, Detlef (Ed.) ; Gesellschaft für Informatik -GI-, Fachgruppe Biometrik und Elektronische Signaturen:
BIOSIG 2008, Special Interest Group on Biometrics and Electronic Signatures. Proceedings : 11.-12. September 2008 in Darmstadt, Germany
Bonn: GI, 2008 (GI-Edition - Lecture Notes in Informatics (LNI) - Proceedings 137)
ISBN: 978-3-88579-231-4
ISSN: 1617-5468
Gesellschaft für Informatik, Special Interest Group on Biometrics and Electronic Signatures (BIOSIG Annual Conference) <6, 2008, Darmstadt>
Fraunhofer IGD ()
biometric; conformance testing; face recognition; ISO/IEC

Digital face images and fingerprint images are included in all newly issued EU passports. Face images stored within electronic passports (ePass) have to conform to ISO/IEC 19794-5 which defines a number of quality requirements for facial image data. A software application (QS Software) is used to verify photographic image properties (pose, head/image size, height-to-page ratio, etc.) as well as setup properties (color depth, contrast, brightness, etc.). To ensure reliability and reproducibility of automated facial image processing systems it is necessary to assess the quality of existing software packages. To this end a database of facial image data would be of significant value as a testing base.
The goal for project series FIReBIRD (Facial Image REcognition Benchmark Including Realistic Disturbances) is to establish a database of facial images that can be used worldwide for compliance and performance testing and quality verification based on ISO standards. FIReBIRD will be conducted in close cooperation of Fraunhofer IGD and the German Federal Office for Information Security (BSI) and will be carried out in parallel and in international cooperation with the U.S. National Institute of Standards and Technology (NIST) and the British National Physical Laboratory (NPL). After successful completion of a conceptual framework the international database will be built. This facial image database will consist of distinct data sets used by two separate user groups for two different purposes: public test data that is available for algorithm developers and non-public test data for system evaluation accessible only to authorized organizations such as BSI/Fraunhofer IGD, NPL and NIST.