Public code of software Robust Curvelet IQA described in article:
Robust statistics and no-reference image quality assessment in Curvelet domain.
Published in XIV Workshop de Visão Computacional (WVC 2018), Ilhéus BA, Brazil.
It is available for consultation at,
http://www.wvc2018.com.br/proceedings
arxiv no double blind version at,
https://arxiv.org/abs/1902.03842
Because 'pyct' this software run under python 2.7;
List of packeges to run Rcurvelet_Features: skimage, sklearn, pandas, numpy, sys, scipy and pyct.
Only 'pyct' don't install by pip install, easy install or anaconda repo.
To install pyct use the repository: https://github.com/slimgroup/PyCurvelab
To make a train features type:
$ python RCurvelet_Features.py path/to/input_image.jpg path/to/output_file.csv
To make a test features:
$ python RCurvelet_Features.py path/to/input_image.jpg path/to/output_file.csv image_class image_survey_score
Watch out for models in models folder.
This module uses trained models. The models were trained with classes jpge, jp2k, Gaussian white noise and Gaussin Blur of data sets Live IQA, TID2013 and CSIQ.
The name of the models follows a fixed structure: a) Regressor models name is: regressor_class_DEGRADATION_model.pkl b) Scale models name is: scale_class_DEGRADATION_model.pkl
To make a score:
$ python RCurvelet_Score.py input.csv #To show
$ python RCurvelet_Score.py input.csv score.csv #To save output file.
Note: input.csv can be a multiline file.
Will appear soon
Use a shell script "usefull_script.sh" to make a scores for a batch of images.
Robust Curvelet IQA 2018.01 For the versions available, see the https://github.com/rgiostri/robustcurvelet/
Ramón Giostri Campos
This project is licensed under terms of the GNU General Public License version 3 - see the [LICENSE.md] file for details
For the professor Evandro Ottoni Teatini Salles for guidance in the works.