CGAN网络在去除图像雨雪痕迹中的应用

Application of CGAN network for rain and snow removing in images

为应对极端天气对码头集装箱图像处理的影响,提高图像算法的识别率,采用条件生成式对抗网络(CGAN)算法去除图像中的雨雪痕迹,将欧几里得损失和改进的感知损失函数方法相结合,以提高生成图像的效果.试验结果表明,CGAN网络算法能够在保留图像特征信息的基础上,有效去除图像中的雨雪痕迹.

In order to deal with the influence of extreme weather on the processing of container number images and improve the recognition rate of image algorithm,the conditional generation adversarial network(CGAN)algorithm is used to remove the rain and snow in image.The method combines Euclidean loss and refined perceptual loss to improve the effect of generating images.Experimental results show that the CGAN network can effectively remove the rain and snow in the image while retaining the image feature information.