U-Net Based Deep Regression Network Architecture for Single Image Super Resolution of License Plate Image

License plate is automatically recognition for various purpose by private or government organization in the universe. As a result of the license plate becoming dull with age, low-quality images are captured for identification. As a result, a deep learning architecture for the high-resolution license plate images was given in this study. Converting low-quality images into high-quality images with the required edge components, including textural information, was the objective of Single Image Super Resolution (SISR). Extra information from the HR pictures is available for use in security, medical imaging, and other applications. It is difficult to get greater precision with less error, because the reconstructed image lacks some information in certain areas. In order to overcome these difficulties, LR pictures are transformed into HR images using a regression network-based super resolution (RNSR). A 50-layer deep regression network is constructed for the SISR. Simulation analysis shows that the proposed RNSR approach converts LR multimedia text images to HR images with 98% accuracy, 94% specificity, and 97% precision. Regression network output can provide images of excellent quality based on the efficiency provided by the proposed RNSR methodology.