Novel Approach of Face Identification Based on Multi-scale Local Binary Pattern

Face recognition is one of the most successful application of biometric system and image analysis. In this work, a novel method for face recognition using the Multi-scale Local Binary Pattern (MLBP) descriptors is proposed. The proposed method consists of two steps as follows: first, feature extraction is applied to extract and compute the MLBP features from face images, second, a matching process for decision making is then applied based on the assessment of the similarities between the feature vectors using the city block distance. The proposed method is evaluated on three popular databases which are: ORL, AR and Yale. The experimental results show its performances compared to similar and recent existing methods.