Face Clustering Based on Fusion of Face Tracking and Optimization.

Face clustering algorithms play an important role in face recognition and data analysis. Since the slow speed and high complexity face clustering algorithm of large-scale data in the face recognition system, we propose a face clustering algorithm combining face tracking and optimization. According to the face image and coordinates obtained by the face detection algorithm, we create face tracks. Then using the face quality assessment algorithm to select the face image with better facial posture and finer image from one of those face tracks and as the representative picture of the face track. Combining the similarity matrix of the face representative image and constraint matrix between face tracks, we can use unsupervised clustering algorithm to cluster representative images of those faces and get better clustering results. Experimental results show that our method can quickly and effectively cluster face images compared with other algorithms.