Face Recognition using Original and Symmetrical Face Images

The biggest bottleneck of face recognition is the insufficient training samples of face images. In reality, face possess varying nature of expressions, illumination, poses, etc. There is a constrained of insufficient training samples due to number of reasons which reduces the accuracy of face recognition techniques. A method to generate various symmetrical face images by exploiting the axis-symmetrical structure of the face is proposed. This method is simple and computationally efficient. The new generated symmetrical samples i.e. Left symmetrical face, Right symmetrical face and Mirror symmetrical face, is basically different from the original face but reflects different appearance of the original face. Classification is performed on all these samples using Principal Component Analysis (PCA) algorithm. Score fusion technique is used for ultimate classification result. Experiments shows that proposed method improves the accuracy of the recognition. It might improve other face recognition methods too.