A proposed framework for robust face identification system

Abstract

Human face is the most representative part of body that can be used to differentiate one person among others. Accurate face identification system is still a challenge to Image Processing and Pattern Recognition researchers. In this paper, a complete framework for face-based personal identification system is proposed. The proposed frame work is composite of three basic stages; face skin detection (FSD), facial features positioning (FFP), representative features extraction (RFE) and face matching (FM). For FSD stage, RGB-H-CbCr color model is used after a comparative study between different color models. Enhanced Haar-like features are utilized for FFP stage. After accurate features positioning, the representative features are calculated using the centers of eyes, nose and mouth organs. The experimental results of this paper depict that the proposed frame work accurately identify persons of The Center for Vital Longevity Face Database. The proposed system could Identify the correct person with 40 saved image with accuracy 98%, while it could reject wrong persons with accuracy 98.17%. The overall accuracy of correct identification reaches 98.14%.