Combined Approach to Face Detection for Video-Surveillance

The paper describes some face detection algorithms using motion detection, skin color segmentation, Haar-like features and neural networks. The face candidate selection stages label promising input image areas that may contain faces. Then, the cascade of the simplified retinally connected neural networks' ensembles performs the final classification of the rest image windows using improved face search strategy across scale and position. The cascade allows speed image processing using small and fast neural networks. The proposed search strategy applies inverse image scale pyramid, adaptive scanning step and window acceptance to decrease the number of windows which should be processed by the classifier.