Evaluation of Image Support Resolution Deep Learning Technique based on PSNR Value

Abstract

Machine learning is a subset of deep learning. Deep learning techniques have become increasingly popular in recent years for a variety of applications. Many techniques are used in image processing for image enhancement, object detection, and a variety of other purposes. Many deep learning techniques are important in single-image super-resolution (SISR). Many models are used for SISR, and they produce good perceptual and adversarial results. The procedure involves creating a high-resolution (HR) image from a low-resolution (LR) image, including some popular SISR methods: "Super-Resolution with Convolutional Neural Network" (SRCNN), "Super-Resolution Residual Network" (ResNet), "Super-Resolution Generative Adversarial Network" (SRGAN), and "Enhanced Super-Resolution Generative Adversarial Network" (ESRGAN). These deep learning models are layered, with each layer performing a different task for an LR image. These methods convert a single LR image into an HR image. A single LR image yields many high-resolution images. This paper creates a dataset of DSLR images for use with the SISR method. Deep learning technology is used to convert an LR image to an HR image. The HR images are then compared to the DSLR images. The "peak signal-to-noise ratio" (PSNR) and the "Structural Similarity Index" (SSI) are used to compare them (SSIM). When compared to the old method, the proposed method produces identical results. SSIM and PSNR are shown for both old and proposed approaches. Both approaches possess almost comparable results. The higher the PSNR value, the better the image quality. The same is the case with the SSIM value