Abstract
In this paper, we propose a classification-based approach for locating human faces in cluttered images. We extract gradient features from local image and the dimensionality of the feature vector consisting of gradient features and gray scales is reduced by the PCA for the sake of computation efficiency and detection efficacy. The polynomial neural network (PNN) takes as inputs the binomials of the projection of the feature vector onto the subspace learned by PCA and is trained on face and non-face samples to discriminate the two classes of patterns. In order to further improve the detection performance, we incorporate into PNN the distance of feature vector from the subspace (DFFS). The experimental results on testing a large number of images with complex backgrounds demonstrated the efficiency of the proposed method in terms of detection performance and computational cost.
