基于Res2net和金字塔池化的图像去雾算法

Image Dehazing Algorithm Based on Res2net and Pyramid Pooling

在计算机视觉的高级任务中,对图像的清晰度有很高的要求,目前基于深度学习的图像去雾算法仍存在一些问题,如细节丢失、色彩失真、去雾不完全等。为解决这些问题,设计了一种基于Res2net和金字塔池化的端到端图像去雾算法。该网络中,通过使用Res2net模块提取上下文特征,并利用金字塔池化模块融合不同尺度的特征信息。为了得到更好的网络模型,采用RESIDE数据集对提出的模型分别进行训练和测试。结果表明:该模型在主客观评价中都取得了不错的效果,极大地改善了去雾后图片色彩失真和去雾不够彻底的问题。

Due to the advanced tasks of computer vision have extremely high requests for the sharpness of the image,but the existing image dehazing algorithms based on deep learning have some problems,such as loss of details,color distortion,and incomplete dehazing.In order to solve these problems,in this paper propose an end-to-end image dehazing algorithm based on Res2net and pyramid pooling to directly obtain clear images.In this network,multi-scale contextual features are extracted by using Res2net,and features at different levels are fused by a pyramid pooling module.In order to get a better network model,this paper uses the RESIDE training set and testing set to train and test the proposed model respectively.The results show that the proposed model achieves good results in both subjective and objective evaluations,and effectively improves the problems of color distortion and incomplete dehazing in dehazing images.