Computer Science Department, Swiss Federal Institute of Technology, Zurich, Switzerland.
This paper exploits a simple but general technique to extract object models from arbitrary image sequences. Such object models can be used to structure and index the image sequence. The algorithm extracts and tracks homogenous regions, which may correspond to objects or object parts. By tracking the absolute as well as the relative position of these regions the algorithm constructs models of potential objects. As such, the approach is model-free in the sense that it does not uses a priori models to detect, track or segment objects. On the contrary, the ultimate goal of the approach is to build such models automatically from image sequences. In this paper the approach is applied to an image sequence taken by a static camera overlooking a parking lot. Since the approach is model-free it can be used to extract any object from image sequences including cars and people. Tracking results for cars and people are reported.