Numerous digital cameras and modern phones have a face detection module, which is used to automatically focus (AF) and optimize exposure (AE). But the face detection will fail when person doesn't face the camera or the part of the face is occluded. In order to avoid such problems, we propose a fast head-shoulder detector, which uses Variable-size block Histograms of Orientated Gradients (VHOG) descriptors. AdaBoost-based feature selection algorithm and integral image representation are used to speed up the algorithm. The tests reveal that the method shows very good results and works efficiently in spite of the low computational power and memory available in mobile devices.
展开▼
