近年来，深度学习的去雾方法在图像去雾领域取得了显著的成绩。然而，大多数基于U型网络的去雾方法将编码层特征直接传递到对应解码层，缺乏浅层和深层特征之间的信息交互。此外，基于非U型网络的去雾方法存在感受野受限问题，无法有效的利用上下文信息。从而导致这些方法在场景尺度变化较大的遥感图像去雾中无法取得理想效果。为此，本文提出了一种基于层级特征交互与增强感受野的双分支遥感图像去雾网络，该方法包含层级特征交互子网和多尺度信息提取子网。其中，层级特征交互子网利用层级特征交互融合模块，逐层的在浅层特征中引入语义信息，深层特征中引入空间细节信息，从而增强编码层中不同层级特征之间的信息交互。多尺度信息提取子网利用多尺度残差空洞卷积模块，融合不同感受野的特征，从而获取对于遥感图像去雾至关重要的上下文信息。在两个公开数据集上的实验结果表明，本文提出的去雾方法相比现有的9种优秀的去雾算法，取得了最好的客观评价指标和视觉效果。

Abstruct: Objective In recent years, deep learning dehazing methods have achieved remarkable results in the field of image dehazing. However, most dehazing methods based on U-shaped networks directly transfer the features of the encoding layer to the corresponding decoding layers, lacking the information interaction between the low-level and high-level features. Meanwhile, the network model designed based on the U-shaped structure may destroy the detailed information important for the restored image in the process of downsampling, so that the restored clear image lacks detailed texture and structure information. In addition, the dehazing method based on non-U-shaped network has the problem of limited receptive field, which can not effectively utilize the context information. As a result, these methods can not achieve ideal dehazing results in remote sensing image with large scene scale changes. Method Therefore, this paper proposes a two-branch remote sensing image dehazing network based on hierarchical feature interaction and enhanced receptive field, which includes hierarchical feature interaction sub-net and multi-scale information extraction sub-net. The hierarchical feature interaction sub-net uses the hierarchical feature interaction fusion module to introduce semantic information into low-level features and spatial details into high-level features layer by layer, thereby enhancing the information interaction between features at different levels in the encoding layer. The multi-scale information extraction sub-net uses the multi-scale residual dilated convolution module to fuse the features of different receptive fields, so as to obtain the context information which is crucial for remote sensing image dehazing. Result The experiment on two public datasets show that the dehazing method proposed in this paper achieves the best evaluation compared to the existing 9 excellent dehazing algorithms. Among them, in the three sub-test sets of the public remote sensing dataset Haze1k, the quantitative index PSNR of this paper reaches 27.362dB, 28.171dB, and 25.137dB. In the two sub-test sets of the public remote sensing dataset RICE, the quantitative index PSNR of this paper reaches 37.792dB and 35.367dB. In addition, the method proposed in this paper is the closest to ground truth in terms of subjective vision such as color, saturation and sharpness while achieving the dehazing effect. Conclusion We can draw the following conclusions:(1) Through the proposed hierarchical feature interaction fusion module, the deep semantic information in the coding stage is gradually interactively fused with the shallow detailed texture information, so as to enhance the expressive ability of the network and restore clear images with higher quality. (2) through the multi-scale residual dilated convolution module, the dehazing network proposed in this paper can increase the receptive field of the network without changing the size of the feature map, and fuse the context information of different scales.(3)In Haze1k and RICE, two public remote sensing image dehazing datasets, the dehazing method proposed in this paper is superior to 9 excellent dehazing algorithms recently proposed in terms of objective evaluation indexes and subjective visual effects.