Combining Contour, Edge and Blob Tracking

C. Town and D. Pugh (UK)


Tracking, Edge Detection, Active Contours, Segmentation.


Most modern approaches to object tracking rely on para metric or non-parametric representations such as edges, active contours, colour blobs, or segmented regions. We present a framework which integrates information from all of these approaches in order to reliably track objects in se quences taken by a single static camera and ascribe para metric motion models to such objects. The resulting sys tem combines many of the benefits of the individual meth ods such as self-initialisation, good convergence to ob ject boundaries, and derivation of motion and appearance models which may serve as a basis for high-level activity analysis.

Important Links:

Go Back