Automatic face detection system for street view images

Abstract

As street view images have been widely used in digital maps, detecting and blurring faces in these images for privacy protection is highly needed. In this paper, an automatic face detection system for street view images is constructed. Firstly, image pre-processing is used to increase image's global contrast. Then, head-shoulder detection based on combined-HOG features is applied to locate faces roughly. After that, multi-angle face detection and Bayes face detection based on skin segmentation are used to get faces' locations precisely from the located head-shoulder regions. Experiments conducted show high precision and robustness of the system.