动态感受野特征选择去雾网络

doi:10.16180/j.cnki.issn1007-7820.2023.07.008

摘要/Abstract

摘要：

基于深度学习的去雾模型大多在网络参数固定后,感受野也就随之固定。这导致去雾网络无法针对每个具体的场景采用最优的模式进行去雾,从而造成结果中存在模糊和失真。针对这些问题,文中提出动态感受野特征选择去雾网络。该网络以带有空洞卷积的特征注意力空洞模块为基础组件,并行使用多个空洞率不同的特征注意力空洞模块来提取多尺度特征,并进行动态特征融合,构成动态感受野模块。文中将多个动态感受野模块搭配残差连接组成深度网络,对不同层次的特征进行动态混合,最终解码得到去雾图像。实验结果表明,文中所提算法对室内和室外的合成雾图以及真实含雾图像均具有良好的去雾效果,可以生成清晰、自然的去雾图像。

关键词: 图像去雾, 动态感受野, 多尺度特征, 动态特征融合, 空洞卷积, 自注意力机制, 动态神经网络, 动态参数

Abstract:

Most of the deep-learning based dehazing models have fixed receptive filed after the parameter are fixed. As a result, the dehazing network cannot adopt the optimal mode for dehazing each specific scene, resulting in ambiguity and distortion in the results. In view of these problems, this study proposes a dynamic receptive field feature selection dehazing network. A feature-attention atrous block with atrous convolution is designed as the basic module of the network. Multiple feature attention atrous blocks with different atrous rates are used in parallel to extract multi-scale features. Dynamic feature fusion is performed on these features to form a dynamic receptive field block. Multiple dynamic receptive field blocks are combined with residual connections to form a deep network. The features from different levels are dynamically mixed and decoded to obtain a haze-free image. The experimental results show that the proposed algorithm has a good dehazing performance on indoor, outdoor, and real hazy images, and can generate clear and natural dehazing images.

Key words: image dehazing, dynamic receptive field, multi-scale features, dynamic feature fusion, atrous convolution, self-attention mechanism, dynamic neural network, dynamic parameters

中图分类号:

TP391.41

查俊伟,张洪艳. 动态感受野特征选择去雾网络[J]. 电子科技, 2023, 36(7): 56-63.

ZHA Junwei,ZHANG Hongyan. Dynamic Receptive Field Feature Selection Dehazing Network[J]. Electronic Science and Technology, 2023, 36(7): 56-63.

图/表 8

图1

图2

图3

图4

表1

图5

图6

表2

参考文献 30

[1]	张莹, 刘子龙, 万伟. 基于Faster R-CNN的无人机车辆目标检测[J]. 电子科技, 2021, 34(11):11-20.
	Zhang Ying, Liu Zilong, Wan Wei. UAV vehicle target detection based on faster R-CNN[J]. Electronic Science and Technology, 2021, 34(11):11-20.
[2]	战荫伟, 朱百万, 杨卓. 一种车脸识别算法的研究与应用[J]. 电子科技, 2021, 34(8):1-7.
	Zhan Yinwei, Zhu Baiwan, Yang Zhuo. Research and application of vehicle face recognition algorithm[J]. Electronic Science and Technology, 2021, 34(8):1-7.
[3]	Middleton W E K. Vision through the atmosphere[M]. Berlin:Springer, 1957: 254-287.
[4]	He K M, Sun J, Tang X O. Single image haze removal using dark channel prior[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 33(12):2341-2353. doi: 10.1109/TPAMI.2010.168
[5]	Fattal R. Dehazing using color-lines[J]. ACM Transactions on Graphics, 2015, 34(1):1-14.
[6]	Yuan F N, Zhou Y, Xia X, et al. A confidence prior for image dehazing[J]. Pattern Recognition, 2021, 11(9):1-10. doi: 10.1016/0031-3203(79)90023-2
[7]	Cai B, Xu X, Jia K, et al. DehazeNet: An end-to-end system for single image haze removal[J]. IEEE Transactions on Image Processing, 2016, 25(11):5187-5198. doi: 10.1109/TIP.2016.2598681 pmid: 28873058
[8]	Ren W Q, Liu S, Zhang H, et al. Single image dehazing via multi-scale convolutional neural networks[C]. Amsterdam: Proceedings of the European Conference on Computer Vision, 2016:154-169.
[9]	Li B, Peng X, Wang Z, et al. AOD-Net: All-in-one dehazing network[C].Venice: Proceedings of the IEEE International Conference on Computer Vision, 2017:4780-4788.
[10]	Liu X, Ma Y, Shi Z, et al. GridDehazeNet: Attention-based multi-scale network for image dehazing[C].Seoul: IEEE/CVF International Conference on Computer Vision, 2019:7313-7322.
[11]	Wang T, Zhao L, Huang P C, et al. Haze concentration adaptive network for image dehazing[J]. Neurocomputing, 2021, 43(9):75-85.
[12]	Hong M, Xie Y, Li C, et al. Distilling image dehazing with heterogeneous task imitation[C].Seattle: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 3459-3468.
[13]	Wu H, Qu Y, Lin S, et al. Contrastive learning for compact single image dehazing[C].Nashville: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021:10546-10555.
[14]	Chen L C, Papandreou G, Kokkinos I, et al. DeepLab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834-848. doi: 10.1109/TPAMI.2017.2699184
[15]	Qin X, Wang Z L, Bai Y C, et al. FFA-Net: Feature fusionattention network for single image dehazing[C].New York: Proceedings of the AAAI Conference on Artificial Intelligence, 2020:11908-11915.
[16]	Woo S, Park J, Lee J-Y, et al. CBAM: Convolutional block attention module[C].Munich: The European Conference on Computer Vision, 2018: 3-19.
[17]	Hu J, Shen L, Sun G. Squeeze-and-Excitation networks[C]. Salt Lake City: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018:7132-7141.
[18]	Han Y, Huang G, Song S, et al. Dynamic neural networks: A survey[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021(6):1-7.
[19]	Zeiler M D, Fergus R. Visualizing and understanding convolutional networks[C].Zurich: The European Conference on Computer Vision, 2014: 818-833.
[20]	Li X, Wang W, Hu X, et al. Selective kernel networks[C].Long Beach: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019:510-519.
[21]	Ren S Q, He K M, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis Machine Intelligence, 2016, 39(6):1137-1149. doi: 10.1109/TPAMI.2016.2577031
[22]	Deng J, Dong W, Socher R, et al. Imagenet: A large-scale hierarchical image database[C].Miami: IEEE Conference on Computer Vision and Pattern Recognition, 2009:248-255.
[23]	Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[C].San Diego: International Conference on Learning Representations, 2015: 1-14.
[24]	Li B, Ren W, Fu D, et al. Benchmarking single-image dehazing and beyond[J]. IEEE Transactions on Image Processing, 2019, 28(1): 492-505. doi: 10.1109/TIP.83
[25]	Ancuti C O, Ancuti C, Timofte R, et al. O-HAZE: A dehazing benchmark with real hazy and haze-free outdoor images[C].Salt Lake City: IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2018:867-8678.
[26]	Wang Z, Bovik A C, Sheikh H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004, 13(4): 600-612. doi: 10.1109/tip.2003.819861 pmid: 15376593
[27]	Berman D, Treibitz T, Avidan S. Non-local image dehazing[C].Las Vegas: IEEE Conference on Computer Vision and Pattern Recognition, 2016:1674-1682.
[28]	Ren W, Ma L, Zhang J, et al. Gated fusion network for single imagedehazing[C].Salt Lake City: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018: 3253-3261.
[29]	Qu Y, Chen Y, Huang J, et al. Enhanced Pix2pix dehazing network[C].Long Beach: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019:8152-8160.
[30]	Dong H, Pan J, Xiang L, et al. Multi-Scale boosted dehazing network with dense feature fusion[C].Seattle: IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 2154-2164.

数据集		DCP	NLD	Dehaze Net	AOD- Net	GFN	EPDN	GDN	MSBDN	CFP	HCAN	本文方法
SOTS_INDOOR	PSNR/dB	19.250	17.290	21.140	19.400	22.300	25.060	31.260	32.920	21.040	31.080 0	35.760
SOTS_INDOOR	SSIM	0.848	0.777	0.847	0.839	0.880	0.923	0.984	0.982	0.871	0.968 6	0.987
SOTS_OUTDOOR	PSNR/dB	16.850	5.540	20.290	23.960	21.550	22.570	30.130	31.360	20.670	32.270 0	32.250
SOTS_OUTDOOR	SSIM	0.840	0.221	0.877	0.926	0.844	0.863	0.981	0.976	0.842 5	0.978 0	0.981
O-HAZE	PSNR/dB	14.820	15.310	12.970	18.350	18.160	16.010	21.810	23.430	18.120	20.930 0	23.070
O-HAZE	SSIM	0.561	0.578	0.476	0.663	0.671	0.637	0.732	0.734	0.517	0.705 0	0.740

	实验1	实验2	实验3	实验4
DRB-1				√
DRB	√	√	√
AFS	√	√		√
DFM	√		√	√
PSNR/dB	35.76	35.15	35.29	33.99
SSIM	0.987	0.986	0.986	0.983