Code repository for the paper:
Fast Blue Noise Generation via Unsupervised Learning
Daniele Giunchi*,
Alejandro Sztrajman*,
Anthony Steed
International Joint Conference on Neural Networks (IJCNN), 2022.
Running the script bn_train.py
will train the blue noise neural network model and save it as model128.h5
and model128.json
, where 128 indicates
the resolution of the square grayscale blue noise masks generated by the network.
Run the following line to generate a blue noise texture using the model model128.h5
:
python bn_predict.py model128.h5 --cpu
This will create an output file pred128.png
with the grayscale blue noise mask.
Use the script dither.py
to perform dithering of an image with our generated blue noise:
python dither.py --images "img/meadow1.png" --noises "pred128.png" --bits 1
This uses the blue noise mask in pred128.png
to dither the image meadow1.png
, compressing it to a single bit per color channel,
outputting the file dither_meadow1_pred128_bits4.png
.
If you find our work useful, please cite:
@article{giunchi2022bluenoise,
author={Daniele Giunchi and Alejandro Sztrajman and Anthony Steed},
title = {Fast Blue-Noise Generation via Unsupervised Learning},
booktitle = {International Joint Conference on Neural Networks},
year = {2022}
}