This paper introduces a priority based edge detection of license plate and recognition of characters. We intend to extract and identify vehicle registration numbers from real time vehicle images without any human intrusion. In computation of license plate, localization shows better edge detection result when compared to conventional methods such as binarization, template matching or transition mapping. Although several effective image processing algorithms are created nowadays, real time operation is found to be impractical as their application is limited by huge computation requirements. As processing time and illumination of the image affects real time operation, it is reduced by setting a dynamic threshold. This paper uses the concept of localization for license plate extraction from real time images followed by the segmentation of characters and reformation.