Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Deraindrop related work #15

Open
meton-robean opened this issue Oct 31, 2019 · 1 comment
Open

Deraindrop related work #15

meton-robean opened this issue Oct 31, 2019 · 1 comment

Comments

@meton-robean
Copy link
Owner

meton-robean commented Oct 31, 2019

早期传统方法: 利用雨滴的物理特性进行建模与定位雨滴位置,使用多帧的上下文信息来去雨滴

  1. H. Kurihata. Rainy weather recognition from in-vehicle camera images
    使用PCA从raindrop patches 中提取主特征, 识别阶段用主特征对测试patch进行模板匹配

  2. Roser and Geiger Video-based raindrop detection for improved image registration ICCV workshops 2009
    作者假设将雨滴建模成球形区域,利用这点去匹配patch中类似的区域来达到识别雨滴的位置

3.Yamashita et al. A Virtual Wiper - Restoration of Deteriorated Images by Using Multiple Cameras
融合多个摄像头所拍到的多张图片的结果来去除雨滴,需要硬件系统支持。

4.Shaodi You Adherent Raindrop Modeling, Detectionand Removal in Video
利用optic flow 和 intensity change 信息来检测雨滴,并且利用视频的前后帧信息来恢复图像。

Repository owner locked and limited conversation to collaborators Oct 31, 2019
@meton-robean
Copy link
Owner Author

meton-robean commented Oct 31, 2019

deraindrop 深度学习类
1.Eigen et al. Restoring an image taken through a window covered with dirt or rain.
使用简单的三层CNN来取出图片中的dirt spots and raindrops. 只能去除小的雨滴,而且结果比较模糊。

  1. Qian. AttentiveGAN Attentive generative adversarial network for raindrop removal from a single image.
    使用了GAN,并且利用注意力机制学习到雨滴的位置信息的heatmap从而提高去雨滴效果。

  2. pix2pix 在image-to-image translation 受欢迎,在qian的工作中也被用来学习有雨到无雨的mapping,但是直接使用效果不理想。

4.Quan et al. Deep Learning for Seeing Through Window With Raindrops ICCV2019
那篇作者认为Qian 用 gt与rain img直接作差所得的binary mask来作为监督信息太简单。 然后他用比较复杂的建模( geometric properties)来获得雨滴的Shape-driven attention map 。再加上channel-wise attetion (关注feature map通道重要性权值)然后做了个双attention的modula插在了每个resblock的中间

5.Liu Dual Residual Networks Leveraging the Potential of Paired Operations
for Image Restoration

作者提出了一种通用的图像复原模块叫 Dual Residual Block (DuRB), 他有四种变体设计可以适用于五类图像复原任务(去雨线、去雨滴,去抖动,去雾等),使用将DuRB插入到encoder-decoder网络进行去雨滴达到了基线效果。

5.还有一个在YUV空间去雨滴的

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Projects
None yet
Development

No branches or pull requests

1 participant