Physics Based Image Deshadowing Using Local Linear Model

Abstract

Image deshadowing algorithms remove shadows from images. This requires both detecting where the shadow is and, once detected, removing it from the image. This work focuses on the shadow removal part. We follow a common physical shadow formation model and learn its parameters using a deep neural network. Our model consists of an existing network for shadow detection, and a novel network for shadow removal. The shadow removal network gets the predicted mask of the shadow region and the shadow image and predicts six parameters per pixel. Remarkably, a straightforward network architecture, that is considerably smaller compared to alternative methods, produces better results on standard datasets1.

Cite

Text

Einy et al. "Physics Based Image Deshadowing Using Local Linear Model." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2022. doi:10.1109/CVPRW56347.2022.00340

Markdown

[Einy et al. "Physics Based Image Deshadowing Using Local Linear Model." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2022.](https://mlanthology.org/cvprw/2022/einy2022cvprw-physics/) doi:10.1109/CVPRW56347.2022.00340

BibTeX

@inproceedings{einy2022cvprw-physics,
  title     = {{Physics Based Image Deshadowing Using Local Linear Model}},
  author    = {Einy, Tamir and Immer, Efrat and Vered, Gilad and Avidan, Shai},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2022},
  pages     = {3011-3019},
  doi       = {10.1109/CVPRW56347.2022.00340},
  url       = {https://mlanthology.org/cvprw/2022/einy2022cvprw-physics/}
}