Non-parametric human segmentation using support vector machine

Kyuwon Kim, Changjae Oh, Kwanghoon Sohn

Abstract: Human segmentation is an important task in digital cameras. In this study, we present a framework of non-parametric human segmentation based on SVM. By exploiting spatial and color features of training images, the framework achieves noticeably better human segmentation results than GrabCut in terms of the overlap ratio with ground-truth.