Non Linear Fusion of Colors to Face Authentication using LDA

摘要

In this article, we propose the use of information color to authenticate face; several spaces of colors were used for the transformation of colorimetric components RGB of the original images. The results obtained in different spaces/or component colorimetric are combined by the use of a nonlinear fusion with a simple neuron network type MLP (Multi layer perceptron). We applied the method of linear discriminant analysis (LDA) to extract the characteristic vector of the face image. To validate this work we tested this approach on frontal images of the data base XM2VTS according to its associated protocol (protocol of Lausanne).