The authentication systems of face generally usedthe grayscale face image as input, but in this paper we studiedthe contribution of the color to the authentication system offace. For the extraction of face characteristics for the database, we tested different spaces colors on the Enhanced Fisherlinear discriminant Model (EFM) which is presented as analternative features extraction algorithm to PrincipalComponent Analysis (PCA) widely used in automatic facerecognition. And once the characteristic vector is extracted,the next stage consists of comparing it with the vectorcharacteristic of face which is authenticated, and with the useof each component color alone at the input of this system, wecalculated the error rates in the two sets of validation and testfor the data base XM2VTS according to the protocol ofLausanne. Finally, the results obtained in different spaces orcomponents colorimetric are combined by the use of anonlinear fusion with a simple neuron network MLP (Multilayer perceptron), the results obtained confirm the efficient ofcolor to improve the performance of an authentication systemof face.
展开▼
