A unified framework for face recognition by using sparse representation

Face recognition algorithms recognize faces well under more constrained environments. They do not achieve higher recognition rate in less controlled environments. The issues of illumination, alignment, pose and occlusion in recognizing face images is crucial when they are dealt simultaneously. We intend to propose a new framework for practical face recognition system that resolves the face recognition issues efficiently in a highly unified manner. The system uses sparse representation tools to handle misalignment and occlusion related problems in face images. Thus, a unified approach for face alignment and recognition in the presence of contiguous occlusion will be achieved. The efficiency of the system is evaluated with the FERET and Yale databases.