Using Bayesian statistics and Gabor Wavelets for recognition of human faces

In this paper, we present a novel approach for recognition of human faces using Markov Random Fields (MRF) and Bayesian models. We examine the relationship between feature vectors in a close proximity system. The feature vectors are coefficients of the 2D Gabor Wavelet Transform (DWGT). The MRF is implemented to match the constraint configurations between the feature vectors. The MRFs posterior probability is formulated to evaluate the MRF configuration for matching constraints between the feature vectors in the query and the test image. The best match is classified using the maximum-a-posteriori (MAP) solution using Mahalanobis distance metrics. The consequent Maximum-A-Posteriori resultant is the expected similarity score for the two face images using ESOP minimization algorithm.