Face detection by means of complex wavelet transforms and neural networks

In this paper a new approach for face detection using two different wavelets transforms, namely, Gabor Wavelets Transform and Dual-Tree Wavelets Transform as feature extractors are proposed. In order to separate face and non-face regions, the Feedforward neural networks have been utilized for classification stage. Two neural networks were trained using the feature vectors obtained from the two wavelet transforms. Both systems, called Gabor+NN and Dual-Tree+NN, showed a competitive detection results. In order to improve the performance of the proposed approach, the detection outputs from both systems were fused into a new system, which is called Gabor_Dual-Tree+NN. Many experimental simulations were carried out on various databases. The face detection results after fusion outperformed the results of Gabor+NN and Dual-Tree+NN with less false detections. The proposed approach has very competitive results with one drawback which is the huge processing time.