A Rule Based Model for Efficient Representation and Accurate Recognition of Human Faces

In this paper, we investigate whether the rule based system derived from edges are effective and efficient for face representation and recognition when the number of classes is fixed and known. The characteristics of edges namely straightness and crookedness are used to derive rules. During training, a representative feature value is calculated for each class based on the average percentage of edges that satisfy the rule. In recognition, the representative feature value of an unknown face image is compared against the class representative feature value to assign a label to it. The proposed system does not require extensive feature extraction and classification techniques. Our experimental analysis on the standard CALTECH face databases shows that the edge based system is suitable for real time environment