Abstract
This paper introduces human-face recognition using edge-based rasterstereography, a biometric identification technique employing pattern recognition, which has the potential to be used as identification-based-security system. A raster grid was projected on human face, distortions of which were used to extract curvature information — unique for a face (principal identification and recognition criterion). This information, in coördinate form, was processed by image-processing algorithms. A modified-raster grid was constructed and tested. The modification, mainly, consisted of use of green dots instead of conventional screen composed of stacked squares, which gave acceptable results for green color. Registration, identi-fication and verification were the three phases of facerecognition system proposed by the authors. In the first step, subject’s face was captured using a digital camera and stored in the system database. In the second one, the registered face was compared with all the stored faces in the database. The verification phase consisted of comparing the identified face with own stored template. Once the facial image passed these three stages, the face was recognized as the one belonging to a particular individual. Our face-recognition system correctly recognized 190 out of 200 human faces during test runs. This algorithm was computationally efficient and effective. The edge-based-rasterstereographic-facerecognition system seemed to be suitable for identification and security purposes as time and cost-saving mechanism.
