A. Elgammal and L. S. Davis. Probabilistic Framework for Segmenting People Under Occlusion. In Proc. of IEEE ICCV, 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....input frame (a) and its foreground from standard background subtraction (b) are used to locate the positions of the heads. 16] also employs an iterative processing to handle the case when the heads are not on the foreground boundary, assuming that the occlusion from each other is slight. In [3], humans are assumed to be isolated as they enter the scene so that a human specific color model can be initialized for segmentation when occlusion occurs. 16] and [11] also use the initialized human specific model to help in tracking through occlusion. All of the above techniques either are ....

A. M. Elgammal and L. S. Davis, Probabilistic Framework for Segmenting People under Occlusion, Proc. Int. Conf. on Computer Vision, Vancouver, Canada, 2001.

....one to overcome the ambiguous and noisy data. The future work may combine other cues such as motion and color to resolve the inherent uncertainty with contour and contour detection errors. The combination of the color and spatial information has been successfully used to segment people in a group [76]. Figure 6.1: Duck Rabbit example of the uncertainty with contour Sixth, the RCR algorithm requires an initial contour extraction which is not available in some cases such as photo analysis. One way to overcome this limitation is to use a face detection system to locate the faces of people in ....

A.M. Elgammal, L.S. Davis, "Probabilistic Framework for Segmenting People Under Occlusion," Proc. Int'l Conf. on Computer Vision, 2001.

....case, using the knowledge of human head position and unexplained foreground ( 11] 7] or vertical projection of the blob ( 4] is able to solve the segmentation problem satisfactorily in most cases. Besides, a human specific model initialized before occlusion can also help solve the problem ([3] [11] 1 This research was supported, in part, by the Advanced Research and Development Activity of the U.S. Government under contract No. MDA 90800 C 0036. Figure 1: A sample input frame (a) and its foreground from standard background subtraction (b) # High density: Blobs consist of large ....

....example frame and its foreground are shown in Fig.1) In this case, many of the heads are not on the foreground boundary and the vertical projection of the big blob is also not informative enough to perform segmentation. Previous research on segmenting and tracking of multiple humans ( 8] 4] 6] [3] [11] 7] etc) has been focused mainly on the first two classes. The high density case is interesting because such scenes usually contain rich human behaviors of interests. However the challenge is also obvious. Color segmentation is not likely to segment the individual humans; motion ....

A. M. Elgammal and L. S. Davis, Probabilistic Framework for Segmenting People under Occlusion, Proc. of Int. Conf. on Computer Vision, Vancouver, Canada, 2001. (a) (c)

No context found.

Ahmed Elgammal and Larry S. Davis, "Probabilistic framework for segmenting people under occlusion, " in Proc. of IEEE 8th International Conference on Computer Vision, 2001.

....segmentation. Bottom left: blob segmentation. Bottom right: constructed occlusion model. a sequence for two targets being tracked throughout occlusion. The part segmentation results are shown as well as the constructed occlusion model. More details and more experimental results can be found in [38]. V. CONCLUSION In this paper, we presented nonparametric kernel density estimation techniques as a tool for constructing statistical representations for the scene background and foreground regions in video surveillance. Since the pdf associated with the Fig. 12. Example results. Top: original ....

A. Elgammal and L. S. Davis, "Probabilistic framework for segmenting people under occlusion," Proc. IEEE 8th Int. Conf. Computer Vision, vol. 2, pp. 145--152, 2001.

.... the same color distribution everywhere inside the blob and since the vertical location of the blob is independent of the horizontal axis, the joint distribution of pixel (x; y; c) the probability of observing color c at location (x; y) given blob A) is a multiplication of three density functions [16]. PA (x; y; c) fA (x)g A (y)h A (c) where hA (c) is the color density of blob A and the densities g A (y) f A (x) represents the vertical and horizontal location of the blob respectively. Estimates for the color density hA (c) can be calculated using the kernel density estimation as was shown ....

....pose. 16 Figure 7 illustrates some blob segmentation examples for various people. Notice that the segmentation and separator detection is robust even under partial occlusion of the target as in the rightmost result. Also in some of these examples the clothes are not of a uniform color. In [16] we showed how this representation can be used to segment foreground regions corresponding to multiple people in occlusion. 4 People Tracking Application In this application we show how the fast Gauss transform algorithm can be used to eciently compute estimate for the gradient of the density as ....

A. Elgammal and L. S. Davis, "Probabilistic framework for segmenting people under occlusion," in 8th IEEE International Conference on Computer Vision, 2001.

....Figure 2 illustrates some blob segmentation examples for various people. Notice that the segmentation and separator detection is robust even under partial occlusion of the target as in the rightmost result. Also, in some of these examples the clothes are not of a uniform color. In [4] we showed how this representation can be used to segment foreground regions corresponding to multiple people in occlusion. The segmentation is achieved by searching for the best arrangement for the people, in terms of 2D translation, that maximizes the likelihood of the foreground. In this ....

....of the foreground. In this application the computation of color probabilities corresponding to different blobs is performed once each frame for each foreground pixel. The search for the best arrangement does not involve re computation of these probabilities. For more details refer to [4]. 5 Experimental Results In this section we present some experimental results that show the speed up that can be achieved using the FGT algorithm for both color modeling applications and for general kernel density estimation. The first experiment compares the performance of the FGT with ....

A. Elgammal and L. S. Davis. Probabilistic framework for segmenting people under occlusion. In Proc. of IEEE 8th International Conference on Computer Vision, 2001.

....segmentation Figure 2 illustrates some blob segmentation examples for various people. Notice that the segmentation and separator detection is robust even under partial occlusion of the target as in the rightmost result. Also, in some of these examples the clothes are not of a uniform color. In [4] we showed how this representation can be used to segment foreground regions corresponding to multiple people in occlusion. The segmentation is achieved by searching for the best arrangement for the people, in terms of 2D translation, that maximizes the likelihood of the foreground. In this ....

....likelihood of the foreground. In this application the computation of color probabilities corresponding to different blobs is performed once each frame for each foreground pixel. The search for the best arrangement does not involve re computation of these probabilities. For more details refer to [4]. 5 Experimental Results In this section we present some experimental results that show the speed up that can be achieved using the FGT algorithm for both color modeling applications and for general kernel density estimation. The first experiment compares the performance of the FGT with ....

A. Elgammal and L. S. Davis. Probabilistic framework for segmenting people under occlusion. In Proc. of IEEE 8th International Conference on Computer Vision, 2001.

No context found.

A. Elgammal and L. S. Davis. Probabilistic Framework for Segmenting People Under Occlusion. In Proc. of IEEE ICCV, 2001.

No context found.

A. Elgammal and L. S. Davis, "Probabilistic framework for segmenting people under occlusion," Proc. IEEE 8th Int. Conf. Computer Vision, vol. 2, pp. 145--152, 2001.

No context found.

A. Elgammal and L. S. Davis, Probabilistic Framework for Segmenting People Under Occlusion, In Proc. of IEEE ICCV, 2001.

No context found.

A. M. Elgammal, and L. S. Davis. Probabilistic framework for segmenting people under occlusion. In Proceeding of the IEEE 8th International conference on computer vision, vol. 2, 2001, pp. 145-152.

No context found.

A. Elgammal and L. S. Davis, Probabilistic Framework for Segmenting People Under Occlusion, IEEE Proc. International Conference on Computer Vision, 2001.

No context found.

A. M. Elgammal and L. S. Davis, "Probabilistic framework for segmenting people under occlusion," in Proceeding of IEEE 8th International Conference on Computer Vision, July 2001, vol. 2, pp. 145--152.

No context found.

A.M. Elgammal, L.S. Davis, "Probabilistic Framework for Segmenting People Under Occlusion," Proc. Int'l Conf. on Computer Vision, 2001.

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