Please login first
Image Enhancement Using Generative Adversarial Networks in Computer Vision
* 1 , 1 , 1 , 1 , * 2, 3
1  Department of Artificial Intelligence, Faculty of Computer Science and Engineering, Ghulam Ishaq Khan Institute of Engineering Sciences and Technology, 23460 Topi, Khyber Pakhtoonkha, Pakistan.
2  Department of Business, University of Europe for Applied Sciences, Think Campus, 14469 Potsdam, Germany.
3  Artificial Intelligence Research (AIR) Group, , Department of Artificial Intelligence, Faculty of Computer Science and Engineering, Ghulam Ishaq Khan Institute of Engineering Sciences and Technology, 23460 Topi, Khyber Pakhtoonkha, Pakistan.
Academic Editor: Eugenio Vocaturo

Abstract:

Image enhancement serves as a critical function in the field of computer vision, improving the quality and clarity of images for various applications. In this study, we present an advanced approach to image enhancement by leveraging the power of Generative Adversarial Networks (GANs). Our method employs a sophisticated GAN architecture, specifically tailored for image enhancement tasks. The GAN model comprises two primary components: a generator and a discriminator. The generator is responsible for producing enhanced images from the input data, while the discriminator evaluates the authenticity of these images, distinguishing between real, high-quality images and the ones generated by the model. Initially, the generator utilizes a deep convolutional neural network (DCNN) to process the input image. It aims to enhance the image by reducing noise, improving resolution, and refining details. The generator is trained to learn the mapping from low-quality images to high-quality counterparts through a series of convolutional and deconvolutional layers, incorporating techniques such as residual learning and attention mechanisms to optimize the enhancement process. Parallelly, the discriminator functions as a binary classifier, assessing the quality of the generated images against real, high-resolution images. The discriminator's feedback is crucial, as it guides the generator to produce more realistic and high-quality images through an adversarial learning process. This dynamic interplay between the generator and discriminator forms the crux of the GAN framework, driving continuous improvement in image quality. Our approach was rigorously evaluated on several challenging datasets, including medical and low-light image datasets. The results underscore the superior performance of our GAN-based method compared to traditional image enhancement techniques. Key achievements include significant improvements in image clarity, reductions in artifacts, and enhanced resolution, all achieved with efficient computational performance. These compelling findings not only validate the effectiveness of our proposed method but also highlight its potential applications in various fields.

Keywords: Image enhancement; Generative Adversarial Networks; Computer Vision; Artificial Intelligence; Convolutional Neural Networks
Comments on this paper
Currently there are no comments available.



 
 
Top