A Comprehensive Analysis to Remove Rain from Single Images Using Generative Adversarial Network

Abstract

A Comprehensive Analysis to Remove Rain from Single Images Using Generative Adversarial Network.”, is a research-based project that main goal is to enhance image quality, improving detection performance and more accurate object detection for single image de-raining processes which removing rain-streak and directly generate a pixel-wise image.

Deraining is a process by which we can get a transparent image by removing raindrops from a rainy image. In the rainy time visibility of any device (Camera) decreases and vision property is affected by the rain. For this reason, capturing images on rainy days reduces the resolution of the images. Currently the security control system, traffic system need exalted-quality images. Obtaining a clear vision during rainy time is an urgent requirement for a better image. At present deraining is making a very nice contribution to removing raindrops from any image. Follows different kinds of generative adversarial networks in the draining process. The common deraining mechanism are GAN. Recently generative adversarial networks are popular. GAN models like Attentive GAN, cGAN, DHSGAN, Cycle GAN which acts as a bridge for decreasing the rain drops. In this paper, we proposed a GAN based method that will reduce raindrops from images and also provide better object detection. For experimenting, we use benchmark dataset consisting of both synthetic and realistic rainy images. We gained outputs that are appreciated both qualitatively and extensively. The DHSGAN GAN provides the greatest representation within that technique.