Enhancing Resolution of Artifact Image using AR-CNN and SRGAN

Abstract

Abstract

The super-resolution strategy recreates a higher-resolution image or arrangement from the observed low resolutions (LR) images. As super-resolution has been created for over three decades, both multi-casing and single-outline super-resolution has critical applications in our everyday life. Existing super-resolution strategies have a few constraints. The artifact is additionally an issue in compression of an image. With artifacts, the high-resolution image is most noticeably terrible to see. In this paper, we try to build up a strategy to make the high-resolution image and expelling artifacts that happen in compression. At first, we use a convolutional neural network that is used to reduce artifacts on images. Then reducing artifacts we send the low resolutions images to super-resolution generative adversarial network to produce high-resolution images. In this way, we get high-resolution images without artifacts. We compare with artifact high-resolution images and without artifact high-resolution images that we get from a super-resolution generative adversarial network (SRGAN). We use a super-resolution generative adversarial network after using artifact reduction CNN (AR-CNN) to get a high-resolution image. We use artifact reduction CNN to remove the artifacts.

Keywords: Super-resolution, Artifacts, Artifact Reduction CNN, Super-resolution Generative Adversarial Network