Using LapSRN (Image Resolution Deep Learning Model) with Transfer Learning

Abstract

In order to create high-resolution photographs, Super Resolution (SR) tries to transform low-resolution photos. SR methods can be categorized into two categories:  Single Image Super Resolution (SISR) and Video Super Resolution (VSR). SISR initially needs to upscale low-resolution photos to high-definition images. VSR, which stands for "image super resolution," is used to transform low-quality videos into ones with higher resolution. Deep learning techniques use Convolutional Neural Networks (CNN), a special sort of deep neural network. Super-resolution images and videos can be processed using a variety of deep learning algorithms. For high-quality image super-resolution reconstruction, CNN are used. Deep Laplacian Pyramid Super-Resolution Network (LapSRN), the current strategy, is based on the CNN SR model. It requires many network parameters and heavy computational loads at run time for generating high-accuracy super resolution results so LapSRN with transfer learning (LapSRN-TL) is proposed. We have analyzed and compared the quantitative and qualitative results of LapSRN-TL with LapSRN deep learning model.