Occluded Face Recognition Using Non-Global Features Extraction and K-Means Clustering Algorithm

Face recognition is a complicated topic with a long history. Nowadays, researchers focus on unconstrained settings such as occlusion, face expression, position movement, and illumination. However, the occlusion issue is receiving less attention in recent research. Many complex deep learning and machine learning techniques have been developed to improve face recognition performance under occlusion conditions. In this paper, we proposed a features extraction technique combining local binary pattern and K-means clustering. The local binary pattern is a two-step algorithm where features aping and encoding are executed. The clustering algorithm converts these features into cluster centers to reduce the vector dimensionality. These feature vectors tested the performance of five classifiers in the matching gallery and probe histogram. Two available datasets, AR for natural occlusion and ORL for synthetic occlusion are employed for the experimental effort.