Investigation of License Plate Character Enhancement Using Super-Resolution Techniques

Abstract

Vehicle license plates contain a variety of information. Therefore, they are used in various systems such as crime prevention and transportation. In this paper, we use the BasicVSR++ model to sharpen vehicle license plates by super-resolution processing. Since the area of the region name is smaller than that of the number, the proposed method improves visibility by weighting the loss function according to the area during training.