Image Raindrop Removal Method for Generative Adversarial Network Based on Difference Learning

Abstract

Abstract Due to the interference of the external environment such as rainy weather on the camera, raindrops can easily adhere to the lens and seriously affect the quality of the photos taken. Therefore, it is of great significance to remove raindrops from the image and improve the quality of the photo. In this paper, a raindrop method for generative adversarial network images based on differential learning is proposed. The general generative network is to input images with raindrops and output clean images. The generative network in this paper does not directly output clean images, but learning the difference between images with raindrops and without raindrops, then subtract the learned difference from the image with raindrops to generate a clean image. In order to learn this difference more effectively, adding reconstruction loss to the generative network, the pre-trained VGG-16 network is used to extract the difference between the generated image and the real image features and calculate the mean square error. The experimental results show, the method in this paper can not only remove the raindrops in the image well, but also reconstruct the image information of the part blocked by the raindrops. The image processed by the algorithm in this paper is tested using the yolov3 target detection algorithm, which can significantly improve the recognition accuracy of the detection algorithm.