Deep Regression Network for the Single Image Super Resolution of Multimedia Text Image
S. Karthick, N. Muthukumaran
Abstract
A key problem in low-level computer vision with several applications is super resolution of a picture. In order to create high resolution (HR) images with the necessary edge structures and texture information, Single Image Super Resolution (SISR) transforms low resolution (LR) photos into HR images. Additional information is provided by HR pictures, which can be used for a variety of tasks like security and medical imaging. However, the rebuilt image suffers from some loss in detail areas, making it more difficult to achieve higher accuracy with lower error. Regression network-based super-resolution (RNSR), which converts LR images into HR images, was created to solve these problems. For SISR, a deep regression network with 50 layers is created. The suggested RNSR approach translates LR multimedia text images into HR images with 98% accuracy, 0.02% error, 97% precision, and 94% specificity, according to the simulation study. The regression network may produce high-quality photos based on the effectiveness of the suggested RNSR approach.