Discriminant color space of maximum scatter difference and face recognition

Color information is important in face recognition. Then what is the optimal face image representation for purpose of recognition? To answer the question, firstly, based on maximum scatter difference discriminant criterion, we found a set of optimal coefficients to combine the R, G and B color components, thus to obtain an optimal color component of original color image. And further, we can obtain three sets of optimal combination coefficients, using which we produce a discriminant color space of 3 dimensions. Experimental results on AR color face database show that the recognition rate in the discriminant color space has been largely improved.