• Stars
    star
    154
  • Rank 242,095 (Top 5 %)
  • Language
    MATLAB
  • License
    Other
  • Created almost 8 years ago
  • Updated over 7 years ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

MATLAB implementation of the paper "Non-Local Image Dehazing", CVPR 2016

Non-Local Image Dehazing

This is the source code implementing the non-local single image dehazing algorithm described in the papers:

Non-Local Image Dehazing. Berman, D. and Treibitz, T. and Avidan S., CVPR2016

Air-light Estimation using Haze-Lines. Berman, D. and Treibitz, T. and Avidan S., ICCP2017

If you use this code, please cite the appropriate paper.

bibtex

@inproceedings{NonLocalImageDehazing,
title={Non-Local Image Dehazing},
author={Berman, D. and Treibitz, T. and Avidan, S.},
booktitle={IEEE Conf. CVPR},
year={2016}, }

@inproceedings{AirlightEstimation,
title={Air-light Estimation Using Haze-Lines},
author={Berman, D. and Treibitz, T. and Avidan, S.},
booktitle={IEEE Conf. ICCP},
year={2017}, }

Getting Started

System Requirements:

The code requires MATLAB, and was tested on windows.

Demo

  1. Change MATLAB's directory to the folder the code was extracted to.
  2. Run the file demo_non_local_dehazing.m in order to see the algorithm's perfomance on the attached images.

Dehazing other images

Each image in the demo is accompanied by a text file containing parameters. These value were taken from the result pages of the paper: "Dehazing using Color-Lines".

The gamma value is required since the image formation model applies to the linear images, where each pixel's value is directly related to the number of photons received at that location on the sensor. A non-linear tone-mapping is applied in camera and most of the images we see are not linear. The gamma value approximates the inverse mapping of the pixels' values to linear.

Despite that, for new images, I recommend trying gamma=1 first (this is the default). It often works even though it is not physically accurate.

Authors

Dana Berman, [email protected]

License

The software code of the non-local image dehazing algorithm is provided for non-commercial use under the attached LICENSE.md