Image defogging method based on deep learning

An image defogging method based on deep learning comprises the following steps: acquiring foggy day image data and sunny day image data, and constructing a sunny-fog pairwise data set; an optimized image defogging Pix2Pix model is constructed; inputting a sunny-fog pairwise data set, and carrying out model training; and inputting a foggy day image to carry out defogging processing. The invention provides a high-resolution network structure, which can process a high-resolution power transmission image, focus on global information by using structural similarity, and can be suitable for relatively complex image information, thereby improving the defogging effect of the power transmission channel image. 一种基于深度学习的图像去雾方法，包括如下步骤：采集雾天图像数据和晴天图像的数据，构建晴雾成对数据集；构建优化的图像去雾Pix2Pix模型；输入晴雾成对数据集，进行模型训练；输入雾天图像进行去雾处理。本发明提出了一种高分辨率的网络结构，能够处理高分辨率的输电图像，利用结构相似性来聚焦于全局信息，可以适用较为复杂的图像信息，从而提升输电通道图像去雾效果。