Version 1.0, 28-Apr-2018

This package contains the MATLAB implementation of "Video Rain Streak Removal By Multiscale Convolutional Sparse Coding".

The code has been tested with MATLAB 2017b on Windows.

Use of this code is free for research purposes only. Shall not be used for commercial purposes! All copyrights belong to the original authors. The technology has applied for patents. If you want to purchase the patents for commercial purposes, please contact the corresponding author: Deyu Meng, [email protected]. Thank you!

================================================================================================

Minghan Li, Qi Xie, Qian Zhao, Wei Wei, Shuhang Gu, Jing Tao, Deyu Meng*, Video Rain Streak Removal By Multiscale Convolutional Sparse Coding. CVPR, 2018.

================================================================================================

1. Unpack the contents of the compressed file to a new directory.

2. Run the DemoY.

================================================================================================

Synthetic rain video:

'park.mat', 'highway.mat', 'man.mat', corresponding to Fig. 3-5 in the reference.

Real rain video: 'wall.mat', 'compfinal.mat', 'night.mat' corresponding to Fig. 6-8 in the reference, respectively.

More experiments on real videos are shown in https://sites.google.com/view/cvpr-anonymity . The dataset can be downloaded from http://gr.xjtu.edu.cn/web/dymeng/2 .

================================================================================================

Code partly borrows from:

[1] Wei Wei, Liyuan Yi, Qi Xie, Qian Zhao, Deyu Meng, Zongben Xu, Should We Encode Rain Streaks in Video as Deterministic or Stochastic? ICCV, 2017.

[2] Shuhang Gu, Deyu Meng, Wangmeng Zuo, Lei Zhang. Joint Convolutional Analysis and Synthesis Sparse Representation for Single Image Layer Separation. ICCV, 2017. Thanks for sharing!

================================================================================================

If you have any questions, please feel free to contact Minghan Li([email protected]).