At present the general shape based detection of traffic signs is an unsolved problem. Approaches that exist either concentrate on a particular shape (e.g. Hough transform for circles) or suffer from a high computational demand (see e.g.[1]). This paper presents a time efficient approach that can be used to detect traffic signs with a shape that can be described as a regular polygon or a circle. The major contribution that leads to a time efficient behavior is the introduction of a twostage detection approach. In the first stage, just the property symmetry is explored to detect candidate regions. In the second step, regular polygons (including circles) can be found.