A Generic Framework for Tracking Using Particle Filter With Dynamic Shape Prior.pdf

1370 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO. 5, MAY 2007 A Generic Framework for Tracking Using Particle Filter With Dynamic Shape Prior Yogesh Rathi, Namrata Vaswani, and Allen Tannenbaum Abstract— Tracking deforming objects involves estimating the shape priors and image statistics for tracking highly deformable global motion of the object and its local deformations as functions objects in the presence of noise and clutter. of time. Tracking algorithms using Kalman filters or particle fil- ters (PFs) have been proposed for tracking such objects, but these A. Past Work have limitations due to the lack of dynamic shape information. In this paper, we propose a novel method based on employing a In order to appreciate this methodology, we briefly review locally linear embedding in order to incorporate dynamic shape information into the particle filtering framework for tracking some previous related work. The possible parameterizations of highly deformable objects in the presence of noise and clutter. The planar shapes described as closed contours are of course very PF also models image statistics such as mean and variance of the important. Various finite-dimensional parameterizations of con- given data which can be useful in obtaining proper separation of tinuous curves have been proposed, perhaps most prominently object and background. the B-spline representation used for a “snake model” as in [2]. Index Terms—Dynamic shape prior, geometric acti