Facial Feature Representation and Face Recognition with Neighborhood-based Binary Patterns

Abstract

This paper presents a study based on the facial feature representation and facial image recognition using the Neighborhood-based Binary Patterns (NBP). Images can be represented as binary code sequences and classified with the related method which is successful in representation of textural features, robust to gray level color changes and can be invariant to the rotation. In proposed study, face images are processed with the simple texture analysis operator of the NBP method and then divided into different number of blocks. Facial feature vectors are generated from the binary code sequences which are obtained by using the intensity values and neighborhood information of the facial image blocks. In order to classify facial images, simple matching coefficient is calculated and K-nearest neighbor (KNN) classifier is performed. Accuracy, precision, recall and F-measure values are measured and compared in performance evaluation tests of the proposed face recognition system. Experimental results observed on the Yale face database show that NBP method has superior performance for representation of facial features and for face recognition than the well-known methods in literature.