J.M. Rehg and T. Kanade, "Visual tracking of self-occluding articulated objects", Technical Report CMU-CS-TR-94-224, Carnegie Mellon Univ. School of Computer Science, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....or motion estimation has been directed at the use of intensity or color images. For example, Aggarwal and Nandhakumar [2] and Huang and Netravali [8] give surveys of several such methods for estimating rigid body motion, Aggarwal et al. 1] review work for estimating articulated motion, and Rehg [14] and Bregler [6] further demonstrate two qualitatively different approaches to tracking articulated motion. These methods implicitly or explicitly calculate from the image data the depth range of some or all of the target s image pixels in determining its 3D pose. If range information is ....

J. Rehg and T. Kanade. Visual tracking of self-occluding articulated objects. Technical Report CMU-CS-94-224, Carnegie Mellon University, 1994.

....( x ) movements in local, joint centered coordinate systems. In another example, Lee and Kunii [22, 23] developed a 27 degree of freedom (DoF) hand skeleton model with an analogous set of constraints. Similar skeleton based models of equal or lesser complexity have been used by other authors [24, 25, 26, 27, 28]. 3.4.2 Appearance Based Model The second group of models is based on appearance of hands arms in the image. This means that the model parameters themselves do not encompass any of the parameters mentioned in Proposition 3 or the ones directly derived from them. They model gestures by relating the ....

....(see [50, 54, 46, 23, 55] for instance) Extraction of fingertip location is then fairly simplified and can be performed using color histogram based techniques. A different way to detect fingertips is to use pattern matching techniques: templates can be images of fingertips [34] or fingers [25] or generic 3D cylindrical models [56] These techniques can be enhanced by using additional image features, like contours [24] Some fingertip extraction algorithms are based on the characteristic properties of fingertips in the image. For instance, curvature of a fingertip follows a ....

J. M. Rehg and T. Kanade, "Visual tracking of self-occluding articulated objects," Tech. Rep. CMU-CS-94-224, Carnegie Mellon University, School of Computer Science, CMU, Pittsburgh, PA 15213, December 1994.

....to be addressed, namely: ffl scale rotation confusion ffl planar rotation ambiguities ffl occlusions ffl implausible model shapes due to linear model The improvements we plan to make to the system are as follows: ffl Address the handling of occlusions. Previous work on this (due to Rehg [15]) has made use of layered templates to model occlusion. We hope to adopt a simpler method whereby we determine the visibility of each vertex individually by considering whether any model facets lie in front of it. ffl Use a non linear modelling technique to improve the accuracy and specificity of ....

J.M. Rehg and T. Kanade. Visual tracking of self-occluding articulated objects. In Proc. ICCV, Boston, MA., 1995.

....to predict a visibility ordering between templates that holds for all bounded object motions between image frames. This section derives general existence conditions for such invariant visibility orders. Specific rules for the on line generation of visibility orders for the hand are described in [11]. 3.1 Binary Occlusion Relations The simplest type of visibility order is a binary occlusion relation between two convex bodies undergoing bounded motion. In binary occlusions, the occluding object A is fully visible, while the occluded object B is obscured. The disjoint relation, A j B, holds ....

....of the fingers and thumb. The combination of kinematic and camera transforms make up a deformation function [12] f (q; s) which maps template coordinates, s = u v] to image coordinates, w = x y] as a function of q. This mapping is illustrated in Fig. 7, and more details can be found in [11]. Template I u v u v Window Function m 1 x y Image f(q,s) 0 Figure 7: A finger tip template and its unit window function. The boundary contour (in white) encloses the template pixels, and the deformation function maps it into the image. The SSD error function measures the difference between the ....

[Article contains additional citation context not shown here]

J. Rehg and T. Kanade. Visual tracking of selfoccluding articulated objects. Technical Report CMU-CS-TR-94-224, Carnegie Mellon Univ. School of Comp. Sci., 1994.

No context found.

J.M. Rehg and T. Kanade, "Visual tracking of self-occluding articulated objects", Technical Report CMU-CS-TR-94-224, Carnegie Mellon Univ. School of Computer Science, 1994.

No context found.

J.M. Rehg and T. Kanade, "Visual tracking of self-occluding articulated objects", Technical Report CMU-CS-TR-94-224, Carnegie Mellon Univ. School of Computer Science, 1994.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC