Multi-Stage Fusion of Local and Global Features Based Classification for Face Recognition

Most existing face recognition approaches have limited performance in uncontrolled environments. Effective face recognition requires several different kinds of feature sets to be taken into account which can integrate heterogeneous and complementary information of the input face images. This paper proposes to fuse two discriminative and complementary feature sets. In this method, Discrete Cosine Transform (DCT) is used to extract global features of facial image while Local Binary Patterns (LBP) are used to extract local descriptors. The classification of these features is performed using the Support Vector Machine (SVM). We have investigated the information fusion both at the feature level and the score level using several combination rules. The proposed fusion approach is validated on FERET face database and has been found to be more reliable than a recognition system which uses only one feature, trained individually.