NH-HAZE

数据集介绍：

这是一个非均匀的真实数据集，具有成对的真实雾度和相应的无雾度图像。这是第一个非齐次图像去模糊数据集，包含55个室外场景。在场景中引入了非均匀雾，使用专业雾发生器模拟雾场景的真实条件。

Access 135+ million publications and connect with 20+ million researchers. Join for free and gain visibility by uploading your research.

IEEE Xplore, delivering full text access to the world's highest quality technical literature in engineering and technology. | IEEE Xplore

The most popular research, guides, news and more in artificial intelligence

NN-HAZE is an image dehazing dataset. Since in many real cases haze is not uniformly distributed NH-HAZE, a non-homogeneous realistic dataset with pairs of real hazy and corresponding haze-free images. This is the first non-homogeneous image dehazing dataset and contains 55 outdoor scenes. The non-homogeneous haze has been introduced in the scene using a professional haze generator that imitates the real conditions of hazy scenes.

Image dehazing is an ill-posed problem that has been extensively studied in the recent years. The objective performance evaluation of the dehazing methods is on

When the rains finally receded early this week, Granite Staters were greeted instead with a haze generated by the wildfires burning out west - a visual reminder that New Hampshire is anecdotally referred to as the “tailpipe” of the country.

基于深度学习的去雾方法如下：基于卷积神经网络的媒介传输估计网络，通过估计模糊图像与其介质传输之间的映射关系实现端到端去雾；不估计传输率和大气光强度，通过轻量级卷积网络直接生成去雾图像；基于平缓卷积扩张组的残差去雾网络，解决了由于扩张卷积块和反卷积造成的图像伪影问题；端对端的特征融合注意网络来直接恢复无雾图像，对合成雾图的去雾效果极好，但对真实雾图去雾效果较差；快速的深层多区域叠加去雾网络，对真实雾天图像的去雾效率较高，但去雾性能较为普通。由于部分算法的雾气数据集为合成雾图，合成方法基于大气散射物理模型，与自然界中真实存在的雾气图在特征分布上有较显著差异，虽然去雾效果优于传统图像增强方法，但对不均匀雾气和浓雾图像去雾效果较差，部分区域易留有残雾。

基于双支残差特征融合去雾方法，分别通过空间域注意分支、通道域注的分支的注意力机制对有雾图像特征进行提取，再对两部分特征进行融合，赋予重要特征的更高的权重，提高有效特征的提取能力，提高有雾图像的去雾效果。