DepthEdgeAwareBENet

A. Karaali, N. Harte, CR. Jung, "Deep Multi-Scale Feature Learning for Defocus Blur Estimation", IEEE Transactions on Image Processing (TIP 2022), 2022. To read paper, please refer: https://arxiv.org/abs/2009.11939

Any papers using this code should cite the paper accordingly.

@ARTICLE{9673106,
  author={Karaali, Ali and Harte, Naomi and Jung, Claudio R.},
  journal={IEEE Transactions on Image Processing}, 
  title={Deep Multi-Scale Feature Learning for Defocus Blur Estimation}, 
  year={2022},
  volume={31},
  number={},
  pages={1097-1106},
  doi={10.1109/TIP.2021.3139243}}

As is, the code produces the results given as the first experimental setting with dataset which is provided in ('Non-parametric blur map regression for depth of field extension'). In order to reach the dataset, you should contact with the author of this paper. The code here includes just 1 sample images from this dataset.

Running

$ python3 BENet.py -i images/image_01.png

In order to make a fair comparisons, we use the same edge maps in some other edge based defocus blur estimations. If you want to use the precomputed edges, which are provided in "recomputed_edges/"

$ python3 BENet.py -i images/image_01.png -e precomputed_edges/edge_01.png 

Please also report any bug to alixkaraali[at_sign]gmail[dot_sign]com