This report describes an efficient methodology of speed sign detection which constitutes of several steps to extract speed signs from frames of video recorded from a web-camera mounted in the front of a car. The most difficult implementation issues that are encountered in this task are the angle of the speed sign, the illumination of the image, the size of the speed sign and the calculation speed. In the proposed detection framework several detection decisions are presented, several of those are conducted by utilizing the k-means clustering algorithm. The detector is evaluated on a database with images recorded from a real scenario. The proposed detection system exhibits relatively good performance with a True Positive Rate (TPr) = 0.9875 and a False Positive Rate (FPr) = 1.03*10^-7 on this database.