An Appearance-Based Prior for Hand Tracking - Advanced Concepts for Intelligent Vision Systems - page 301

is violated. Our improved approach integrates three image cues, making it robust to
incorrect information from two modalities.
We demonstrated improved performance with an appearance-based prior on the
tracking of hands in videos without a static background. As an extension to Flock of
Features tracking it provides a third, orthogonal image cue that the tracking decision
can be based on. This makes it more robust to strong background gradients, back-
ground in a color similar to the hand color, and rapid posture changes. The resulting
tracking method is rather robust and operates indoors and outdoors, with different peo-
ple, and despite dynamic backgrounds and camera motion. The method was developed
for a real-time gesture recognition application and currently requires around 20ms per
720x480 video frame.
Despite the advantages of the chosen appearance-based prior, we are currently eval-
uating the performance of other methods including some for whole objects and some
parts-based approaches [18,10]. The integration of their results into this tracking frame-
work follows the same multimodal fusion approach as this paper's contribution.
While the current interest in virtual and augmented reality as well as 3D technolo-
gies provides ample applications and need for good hand tracking, this method is not
limited to hands but likely also applicable to tracking of other articulated objects such
as pedestrians, for example, for surveillance applications.
References
1. Birchfield, S.: Elliptical head tracking using intensity gradients and color histograms. In:
Proc. IEEE Conference on Computer Vision and Pattern Recognition, pp. 232-237 (June
1998)
2. Bradski, G.R.: Real-time face and object tracking as a component of a perceptual user inter-
face. In: Proc. IEEE Workshop on Applications of Computer Vision, pp. 214-219 (1998)
3. Bretzner, L., Laptev, I., Lindeberg, T.: Hand Gesture Recognition using Multi-Scale Colour
Features, Hierarchical Models and Particle Filtering. In: Proc. IEEE Intl. Conference on Au-
tomatic Face and Gesture Recognition, Washington D.C., pp. 423-428 (2002), http://
citeseer.nj.nec.com/bretzner02hand.html
4. Cutler, R., Turk, M.: View-based Interpretation of Real-time Optical Flow for Gesture Recog-
nition. In: Proc. IEEE Intl. Conference on Automatic Face and Gesture Recognition, pp.
416-421 (April 1998)
5. Dambreville, S., Rathi, Y., Tannenbaum, A.: A Framework for Image Segmentation Using
Shape Models and Kernel Space Shape Priors. IEEE Trans. Pattern Analysis and Machine
Intelligence 30(8) (August 2008)
6. Freund, Y., Schapire, R.E.: A decision-theoretic generalization of online learning and an
application to boosting. In: Vitanyi, P.M.B. (ed.) EuroCOLT 1995. LNCS, vol. 904, pp. 23-
37. Springer, Heidelberg (1995)
7. Isard, M., Blake, A.: A mixed-state CONDENSATION tracker with automatic model-
switching. In: Proc. Intl. Conference on Computer Vision, pp. 107-112 (1998)
8. K olsch, M., Turk, M.: Fast 2D Hand Tracking with Flocks of Features and Multi-
Cue Integration. In: IEEE Workshop on Real-Time Vision for Human-Computer
Interaction, CVPR (2004), http://www.cs.ucsb.edu/%7Ematz/HGI/
KolschTurk2004Fast2DHandTrackingWithFlocksOfFeatures.pdf