Image-based Car License Plate Recognition (CLPR) systems provide an inexpensive automatic solution for remote vehicle identification. Localization stage of the CLPR yields a gray-scale plate clip with printed characters. This paper describes the method of plate clip segmentation into isolated characters, feature extraction and classification. The method is independent on character size, thickness, illumination and is capable of handling plates from various countries. The method uses extensively the gray-scale information and is robust to breaks in character connectivity. It is tolerant to character deformations, such as shear and skew. Promising results have been obtained on Israeli and Bulgarian plates.