Rain removal in single image system using CNN with guided and L0-smoothing filters

Document Type : Research Paper


Department of ECE, PSG College of Technology, Coimbatore, India


In this work, a robust rain removal algorithm is proposed for removing rain from still images. The algorithm uses a deep network architecture called DerainNet for effective rain removal. The proposed network directly learns the mapping relationship between rainy and clean image detail layers from the given set of data. In order to modify the objective function and also to improve the deraining process, other Deep CNN based architecture increases the width or depth of the neurons, which in turn increases the complexity of the network. But this work makes use of the Image Processing domain knowledge which reduces the complexity of the network. Instead of training the entire image, only the detail layer of the image is trained. The detailed layer of the image is obtained using two low-pass filters one after the other. They are guided filter and L0-Smoothing filter. The results obtained prove that the proposed network performs better deraining on images in comparison to paper [2] with light rain streaks. Python version 3.8 is used for this work.


[1] D.Y. Chen, C. Chen and L.W. Kang, Visual depth guided colour image rain streaks removal using Sparse coding,
IEEE Trans. Circ. Syst. Video Technol. 24(8) (2014) 1430-1455.
[2] X. Fu, J. Huang, X. Ding, Y. Liao and J. Paisley, Clearing the skies: A deep network architecture for single image
rain removal, IEEE Trans. Image Proces. 26(6) (2017).
[3] X. Fu, J. Huang, D. Zeng and Y. Huang, Removing rain from single images via a deep detail network, IEEE Conf.
Computer Vision and Pattern Recog. (2017).
[4] Y.H. Fu, L.W. Kang, C.W. Lin and C.T. Hsu, Single frame-based rain removal via image decomposition, IEEE
Int. Conf. Acoustics, Speech and Signal Proces. (2011) 1453–1456.
[5] L.W. Kang, C.W. Lin and Y.H. Fu, Automatic single image-based rain streak removal via image decomposition,
IEEE Trans. Image Proces. 21(4) (2011).
[6] L.W. Kang, C.W. Lin, C.T. Lin and Y.C. Lin, Self-learning-based rain streak removal for image, IEEE Int. Symp.
Circ. Syst. (2012).
[7] J.H. Kim, C. Lee, J.Y. Sim and C.S. Sim, Single image de-raining using an adaptive non local means filter, IEEE
Int. Conf. Image Proces. (2013) 914–917.
[8] P. Li, J. Tian, Y. Tang, G. Wang and C. Wu, Model based deep network for single image de-raining, IEEE Trans.
Image Proces. 26 (2020) 14036–14047.
[9] S.H. Sun, S.P. Fan and Y.F. Wang, Exploiting image structural similarity for single image rain removal, IEEE
Trans. Image Proces. (2014).
[10] R, Qian, R.T. Tan, W. Yang, J. Su and J. Liu, Attentive generative adversarial network for raindrop removal
from a single image, arXiv, CoRR, (2018).
[11] M. Wang, L. Chen, Y. Liang, H. Huang and R. Cai, Deep learning method for rain streaks removal from single
image, J. Engin. 13 (2020) 555–560.
[12] J. Wang, S. Liu, C. Chen and B. Zeng, A hierarchical approach for rain or snow removing in a single colour
image, IEEE Trans. Image Proces. 26(8) (2017).
[13] J. Xu, W. Zhao, P. Liu and W. Tang, Removing rain and snow in a single image using guided filter, IEEE Int.
Conf. Comput. Sci. Autom. Engin. 2 (2012) 304–307.
[14] W. Yang, R. Tan, J. Feng, J. Liu, Z. Guo and S. Yan, Joint rain detection and removal from a single image with
contextualized deep networks, IEEE Trans. Pattern Anal. Machine Intell. 42(6) (2020) 1377–1393.
[15] X. Zheng, Y. Liao, W. Guo, X. Fu and X. Ding, Single image-based rain and snow removal using multi guided
filter, Int. Conf. Neural Inf. Proces. (2013) 258–265.
Volume 12, Special Issue
December 2021
Pages 1679-1691
  • Receive Date: 03 October 2021
  • Revise Date: 08 November 2021
  • Accept Date: 26 November 2021
  • First Publish Date: 26 November 2021