00-036: Gabor Fisher Classifier Face Recognition

The Gabor Fisher Classifier (GFC) is a novel face recognition method which is robust to changes in illumination and facial expression.

Its application is highly relevant to technologies supporting homeland security and is potentially applicable in a wide range of products, especially those offering visual surveillance and Human-Computer Intelligent Interaction.

The GFC method draws upon a large library of pictures from which a set of "primary" faces analogous to the set of primary colors successfully reconstructs any face by "mixing" - a process similar to mixing paint from primary colors.

To identify a face, the method
1. calculates a numerical measure - the "mixture" formula,
2. compares the formulas with faces in the library
3. identifies the image - even when the face has undergone changes such as might occur from differences in illumination.

Market Significance:
Test results from GFC using 600 FERET frontal face images corresponding to 200 subjects with varying facial expressions and lighting conditions showed the Gabor Fisher Classifier achieves 100% accuracy using only 62 features.

The GFC method is effective both in terms of both absolute performance indices and comparative performance against some of the more popular face recognition schemes such as the traditional Eigenfaces method and the Gabor wavelet based classification methods.

Further information about this technology and other research being conducted by the Biometrics Group in the School of Information Technology and Engineering is available at http://cs.gmu.edu/-wechsler/FORENSIC/index.html or at the website for the Center for Distributed and Intelligent Computation http://cs.gmu.edu/-wechsler/DIC _Center/.

US Patent 6,775,415