Photorealistic Style Transfer with Screened Poisson Equation

Roey Mechrez , Eli Shechtman and Lihi Zelnik-Manor. BMVC 2017.

[The Project Page]

SPE - MATLAB 1.0 (2017-02-21) Copyright 2006-2017 Roey Mechrez Licensed for noncommercial research use only.

<div align='center'> <img src='example.png' color="black"> </div>

Background

This code implements a fast post processing to make images photorealistic using Screened Poisson Equation (SPE)

Style transfer algorithms generate unrealistic images, this algorithm is a post-processing to any style transfer method (such as Gatys et al.). The algorithm solves the Screen Poisson equation to combine the stylized image with the original image gradients.

For more information see:

@article{mechrez2017Photorealistic,
  title={Photorealistic Style Transfer with Screened Poisson Equation},
  author={Mechrez, Roey, Shechtman, Eli and Zelnik-Manor, Lihi},
  journal={BMVC},
  year={2017}
}

[arXiv] Please cite these paper if you use this code in an academic publication.

Use

To run one pair of images use

example.m

core functions:

doPoissonCombination.m, updateImagePoisson.m

if you are using Linux or OS, see issue#2 for solution since the attach mex file will not work.

License

This software is provided under the provisions of the Lesser GNU Public License (LGPL). see: http://www.gnu.org/copyleft/lesser.html.

This software can be used only for research purposes, you should cite the aforementioned papers in any resulting publication.

The Software is provided "as is", without warranty of any kind.

Code References

[1] Mechrez, Roey and Shechtman, Eli and Zelnik-Manor, Lihi. "Saliency Driven Image Manipulation". arXiv preprint arXiv:1612.02184. 2016 url

[2] Darabi, Soheil and Shechtman, Eli and Barnes, Connelly and Goldman, Dan B and Sen, Pradeep. "Image melding: Combining inconsistent images using patch-based synthesis." ACM Trans. Graph. 2012 url

[3] Fujun Luan, Sylvain Paris, Eli Shechtman, Kavita Bala. "Deep Photo Style Transfer". CVPR. 2017 url

Version History

• Version 1.0 (2017-02-21) Initial Release