Combining Gradientfaces, principal component analysis, and Fisher linear discriminant for face recognition

There are many difficulties for face recognition, such as varying illumination, facial expressions, high dimensions of images, and optimal classifier design. The facial features would be vague by varying illumination and the positions of facial features would be changed by expressions. These problems influence on the recognition results heavily. Moreover, the dimensions of raw images are very high for computation. Therefore the classifier of face recognition should have the abilities to reduce the image dimensions and classify accurately. In this paper we propose an approach to combine Gradientfaces, principal component analysis (PCA), and Fisher linear discriminant (FLD) for face recognition. And the recognition rate is achieved to 99.39%. This is because we use Gradientfaces to solve the problems from varying illumination and facial expressions. And we combine PCA and FLD to reduce the image dimensions and classify accurately. Therefore, the proposed approach, applying Gradientfaces to PCA and FLD classifier, not only reduces the input dimensions and speeds up the computation time, but also increases recognition rates.