Face detection using one-class SVM in color images
Hongliang Jin, Qingshan Liu, Hanqing Lu, Xiaofeng Tong · 2005
In this paper, a novel face detection approach is proposed in color images, which combines skin color, shape and statistical features. First, an efficient skin color model is used to look for face candidate and then mosaic image based on shape feature is adopted to further reduce candidates. Finally, statistical face model learned by one-class SVM (OCSVM) is performed to verify the rest candidates. It is different from many proposed appearance-based approaches that OCSVM-based face model considers face detection task as a one-class pattern problem and overcomes the problem of hardly representing nonface pattern efficiently. Extensive experiments show that our proposed method has an encouraging performance.