Enhancing Defect Image Through Generative Adversarial Network

Abstract

In recent years, technology has emerging with the modern era emerges. In today’s contemporary world, technology influences in many aspects and the significance of the images has highest priority. This modern technology uses the image for different purposes, which includes the security, identity verification, entertainment, healthcare’s mainly for hospitals, in professional works, and many more. Images can be captured in different devices and in different environments. The assurance that the image taken was very clear and without noise is not provided. And all devices will not work on all different environments. For example, the image taken in the mist or the image of the object in water, was not clear, it may blur or it may contain some noises. In this paper, will address the problem of impure images like low resolution or blur images to enhance the quality through GAN algorithm (Generative adversarial Network algorithm). GAN is important approach in deep learning models. GAN works on the principle of Generative models. The main principle of the GAN (Generative adversarial Network) model is to autonomously identify the similarity or pattern in the input data, and it will try to feasibly resemblance of the original dataset. Through this project, the low resolution or blur images where recovered to clear images.