One of the application areas of the character recognition systems is license plate recognition. This application area has advantages and disadvantages compared to document processing applications. The most significant advantage is the known font characteristics of the license plates. On the other hand, beside the advantages there are many disadvantages of this application area. Especially, acquiring the images from out door environments, physically dirty or damaged plates and the requisite of finding precisely the location of the plate frame, increase the difficulty of the application. In order to over come these difficulties, one of the enhancements is utilizing more than one image frame and using a composition of these frames: for recognition. For this aim, the first task to be accomplished is tracking the frame of the plate through the video sequence. In this work, we present a theoretical and practical solution to the license plate tracking application.