Real-Time Video Dehazing Based on Absorption Transmission Compensation and Spatio-Temporal Guided Image Filtering
CUI Tong1,2,3, TIAN Jiandong1,2, WANG Qiang1,2,3, REN Weihong1,2,3, TANG Yandong1,2
1. State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang 110016, China;
2. Institutes for Robotics and Intelligent Manufacturing, Chinese Academy of Sciences, Shenyang 110016, China;
3. University of Chinese Academy of Sciences, Beijing 100049, China
Abstract：The difficulty of the video dehazing technology is to guarantee the spatial-temporal consistency of the video data. A spatio-temporal guided image filtering (ST-GIF) optimization algorithm is proposed to solve this problem. The spatial and temporal consistencies of the videos' interframe information are taken into consideration in the algorithm. For one thing, the transmission texture is smoothed and the salient boundary is protected. And for another, the flicker noise in the videos are suppressed to ensure the fluency of the haze-free video. Since the classical dehazing model only concerns the influence of scattering on the fog formation, most of the dehazing algorithms based on this model usually generate over-saturated noise in the close shot. A transmission estimation algorithm based on absorption transmission compensation is proposed to solve this problem. This algorithm overcomes the defect that the classical model ignores atmospheric absorption attenuation, significantly improves the accuracy of transmission estimation, and effectively suppresses the over-saturations in the close shot. The proposed algorithm is experimentally compared with several state-of-the-art algorithms on both real-world and synthetic haze video data. The quantitative evaluation results with reference show that the proposed algorithm is at least 12% and 3.4% higher than the others in the two metrics of the signal-to-noise ratio and the structural similarity respectively. As an evaluation method without reference, the visible boundary restoration metric of the proposed algorithm is at least 5.7% higher than the others. Results demonstrate that the proposed real-time video dehazing algorithm can recover the high frequency information more effectively, and improve the image contrast more properly, and the colours of the obtained dehazing videos are more natural and authentic.
[1] Tan R T. Visibility in bad weather from a single image[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA:IEEE, 2008:1-8.
[2] Kim J H, Jang W D, Sim J Y, et al. Optimized contrast enhancement for real-time image and video dehazing[J]. Journal of Visual Communication and Image Representation, 2013, 24(3):410-425.
[3] He K M, Sun J, Tang X O. Guided image filtering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(6):1397-1409.
[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, 2011, 33(12):2341-2353.
[5] Nishino K, Kratz L, Lombardi S. Bayesian defogging[J]. International Journal of Computer Vision, 2012, 98(3):263-278.
[6] Zhu Q S, Mai J M, Shao L. Single image dehazing using color attenuation prior[C]//25th British Machine Vision Conference. 2014. DOI:10.5244/C.28.114.
[7] Meng G F, Wang Y, Duan J Y, et al. Efficient image dehazing with boundary constraint and contextual regularization[C]//IEEE International Conference on Computer Vision. Piscataway, USA:IEEE, 2013:617-624.
[8] Cui T, Tian J D, Wang E N, et al. Single image dehazing by latent region-segmentation based transmission estimation and weighted L1-norm regularization[J]. IET Image Processing, 2016, 11(2):145-154.
[9] Fattal R. Dehazing using color-lines[J]. ACM Transactions on Graphics, 2014, 34(1):1-14.
[10] Berman D, Treibitz T, Avidan S. Non-local image dehazing[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA:IEEE, 2016:1674-1682.
[11] Cai B L, Xu X M, 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.
[12] Ren W, Liu S, Zhang H, et al. Single image dehazing via multi-scale convolutional neural networks[M]//Lecture Notes in Computer Science, Vol.9906. Berlin, Germany:Springer-Verlag, 2016:154-169.
[13] Li B, Peng X, Wang Z, et al. An all-in-one network for dehazing and beyond[A/OL].[2018-05-15]. https://arxiv.org/pdf/1707.06543.pdf.
[14] Lü X Y, Chen W B, Shen I F. Real-time dehazing for image and video[C]//18th Pacific Conference on Computer Graphics and Applications. Piscataway, USA:IEEE, 2010:62-69.
[15] Li Z W, Tan P, Tan R T, et al. Simultaneous video defogging and stereo reconstruction[C]//IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, USA:IEEE, 2015:4988-4997.
[16] Cai B L, Xu X M, Tao D C. Real-time video dehazing based on spatio-temporal MRF[M]//Lecture Notes in Computer Science, Vol.9917. Berlin, Germany:Springer-Verlag, 2016:315-325.
[17] Li B, Peng X, Wang Z, et al. End-to-end united video dehazing and detection[A/OL].[2018-06-01]. https://arxiv.org/pdf/1709.03919.pdf.
[18] McCartney E J. Optics of the atmosphere:Scattering by molecules and particles[J]. Physics Bulletin, 1976, 28(11):421.
[19] Nayar S K, Narasimhan S G. Vision in bad weather[C]//7th IEEE International Conference on Computer Vision. Piscataway, USA:IEEE, 1999:820-827.
[20] McGlamery B L. A computer model for underwater camera systems[C]//Proceedings of SPIE, Vol.208. Bellingham, USA:SPIE, 1980:221-232.
[21] Bouguer P. Essai d'optique sur la gradation de la lumiére[M]. Paris, France:chez Claude Jombert, 1729.
[22] 李耀宇,王宏民,张一帆,等.基于结构化深度学习的单目图像深度估计[J].机器人,2017,39(6):812-819. Li Y Y, Wang H M, Zhang Y F, et al. Structured deep learning based depth estimation from a monocular image[J]. Robot, 2017, 39(6):812-819.
