Accurate Foreground Extraction Using Graph Cut with Trimap Estimation
This paper describes an accurate human silhouette extraction method as applied to video sequences. In computer vision applications that use a static camera, the background subtraction method is one of the most effective ways of extracting human silhouettes. However it is prone to errors so performance of silhouette-based gait and gesture recognition often decreases significantly. In this paper we propose two-step segmentation method: trimap estimation and fine segmentation using a graph cut. We first estimated foreground, background and unknown regions with an acceptable level of confidence. Then, the energy function was identified by focussing on the unknown region, and it was minimized via the graph cut method to achieve optimal segmentation. The proposed algorithm was evaluated with respect to ground truth data and it was shown to produce high quality human silhouettes.
KeywordsGesture Recognition Foreground Object Shadow Detection Unknown Region Background Subtraction Method
Unable to display preview. Download preview PDF.
- 1.Boykov, Y., Jolly, M.: Iterative graph cuts for optimal boundary and region segmentation of objects in N-D Images. In: Proc. IEEE Int. Conf. on Computer Vision, pp. 105–112 (2001)Google Scholar
- 4.Chuang, Y.-Y., Curless, B., Salesin, D., Szeliski, R.: A bayesian approach to digital matting. In: Proc. Int. Conf. Computer Vison and Pattern Recognition 2, pp. 264–271 (2001)Google Scholar
- 6.Horprasert, T., Harwood, D., Davis, L.S.: A statistical approach for real-time robust background subtraction and shadow detection. In: Proc. IEEE Frame Rate Workshop, pp. 1–19 (1999)Google Scholar
- 7.Harville, M.: A framework for high-level feedback to adaptive, per-pixel, mixture-of-Gaussian background models. In: Proc. European Conf. on Computer Vision, pp. 543–560 (2002)Google Scholar
- 8.Kolmogorov, V., Criminisi, A., Blake, A., Cross, G., Rother, C.: Bi-layer segmentation of binocular stereo video. In: Proc. Int. Conf. on Computer Vision and Pattern Recognition (2005)Google Scholar
- 10.Li, H., Greenspan, M.: Multi-scale gesture recognition from time-varying contours. In: Int. Conf. Computer Vision, pp. 236–243 (2005)Google Scholar
- 13.Senior, A.: Tracking people with probabilistic appearance models. In: Proc. IEEE Int. Workshop on PETS, pp. 48–55 (2002)Google Scholar
- 16.Tian, Y.-L., Lu, M., Hampapur, A.: Robust and efficient foreground analysis for real-time video surveillance. In: Proc. Int. Conf. Computer Vision and Pattern Recognition, pp. 970–975 (2005)Google Scholar
- 17.Tu, Z.: An integrated framework for image segmentation and perceptual grouping. In: Int. Conf. Computer Vision, pp. 670–677 (2005)Google Scholar