Obstacle detection in planar worlds is an important part of computer vision because it is indispensable for collision prevention of autonomously navigating moving objects. For example, vehicles driving without human guidance need robust prediction of potential obstacles, like other vehicles or pedestrians. Most common approaches of obstacle detection so far have used analytical and statistical methods like motion estimation or generation of maps. The proposed procedures are mostly composed of many processing steps, so that error propagation of successive steps often leads to inaccurate results. Another problem is the necessity of high computing power for real time applications. In this contribution we demonstrate that obstacle detection in planar worlds can be performed efficiently using cellular neural networks. Beside a fast processing speed the proposed method is also very robust.