A fast method for locating vehicle license plates is proposed in this paper. This approach is primarily based on the observations that the intensity contrast between license plate background color and symbol color is high and that symbol edges are clustered in the plate area. A simple edge detection method based on first derivatives is applied to a car image such that the binary image containing only vertical edges is obtained. Then possible plate regions could be located from the horizontal and vertical projections of the binary image. No preprocessing, such as image enhancement and noise filtering, is required. In addition to its simplicity and fastness, the proposed method is less sensitive to illumination changes of the plate. The proposed method was experimented with 117 diverse car images on a Pentium-133 processor. The system successfully located 116 license plates in 0.2 second on an average.