Nonlinear Fusion of colors to face authentication

Abstract

In this paper, we propose to introduce the color information to authenticate face. To improve the performance of this system, many color spaces have been used for processing RGB color components of the original images. The results in different spaces or components colorimetric are combined by using a nonlinear fusion for classification with networks neurons simple type MLP (Multi layer perceptron). We have applied the method of principal components analysis (PCA) or (Eignenfaces) for the extraction of feature vectors. To validate this work we have tested this approach on front images of the database XM2VTS according to its Associated Protocol (Protocol of Lausanne).