Object Tracking with the Level Set Method and the Particle Filtering

E. Yang and M. Jeon (Korea)


Geometric active contour, Level set method, Particle Filtering, Computer Vision


In this paper, we propose a new approach to tracking an object in a sequence of images, where we combine a particle filtering and a new variational level set method to estimate the position, and to segment the deformed shape of object. A geometric active contour is represented by a level set, and its simple parameterization with centroid is used to track the object. To estimate the global motion of the target as images are obtained sequentially, the particle filtering algorithm is used as the main tool, which can handle the nonlinear and non-Gaussian movement of the object successfully. The main advantage of our method is to reduce the computational complexity, resulting from separating tracking of the deformation of the object from the position estimation, and implementing fast and robust level set method. Our computational experiment demonstrates that the proposed method reduces computation time substantially.

Important Links:

Go Back