A study of face recognition algorithm based on wrapper model and Bayesian decision

In this paper, firstly, a wrapper model is built. According to this model, optimum parameters of Gabor kernel function, and optimal projection vectors of Principal Component Analysis (PCA) are confirmed. Secondly, classifier is obtained on the basis of Bayesian theory, and in this process, a fast method to estimate class-conditional-probability by training samples is presented. Experimental results based on Olivetti Research Laboratory (ORL) Face Database show that, this algorithm can efficiently recognize face images under different illumination, posture and expressions.