Face verification using local binary patterns generic histogram adaptation and Chi-square based decision

We propose a simple yet efficient approach to face authentication using local binary patterns (LBP) and histogram adaptation. First, a generic face model is constructed as the LBP histogram extracted from the face images of different users. Then, MAP adaptation technique is applied to obtain the user-specific LBP histogram using the generic model as a prior. We propose a decision rule based on Chi-square distance for claim verification. We extensively evaluate the proposed approach on two publically available databases (namely XM2VTS and FERET) using standard protocols, and compare the obtained results against those of two related works in the literature. The proposed approach yields in interesting results both in terms of authentication accuracy and computational simplicity.