On video based face recognition through adaptive sparse dictionary

摘要: Sparse representation-based face recognition has gained considerable attention recently due to its robustness against illumination and occlusion. Recognizing faces from videos become a topic of importance alleviate the limit information content in still images. However, sparse framework is not applicable video-based sensitivity towards pose alignment changes. In this paper, we propose method which improves upon representation framework. Our key contribution an intelligent adaptive dictionary that updates current probe image into training matrix based on continuously monitoring video through novel confidence criterion Bayesian inference scheme. Due approach, our robust hence can be used recognize unconstrained successfully. Moreover, moving scene, camera angle, other imaging conditions may change quickly leading performance loss accuracy. such situations, it impractical re-enroll individual re-train classifiers continuous basis. approach addresses these practical issues. Experimental results well known YouTube Face database demonstrates effectiveness method.