Abstract: License plate binarization is a necessary step in the processing license plate recognition,after the license plate has been located and segmented exactly.Generally,Bernsen algorithm,Ostu algorithm and LEVBB algorithm are used for the license plate binarization.However,because of the deficiency in illumination and all kinds of pollution and defile,the above algorithms can not obtain good results.For this,this paper puts forward a new license plate binarization algorithm based on the features of characters' strokes.Quantities of experiments have indicated that the algorithm can overcome the problems and all kinds of pollution,and extract the character target exactly.