REDLOG: Reduced-Reference Quality Assessment Based on the Entropy of DWT Coefficients of Locally Weighted Gradient Magnitudes
S. Alireza Golestaneh, and Lina J. Karam
Here we propose a training-free low cost RRIQA method that requires a very small number of RR features (6 Features at most).
@article{golestaneh2016reduced,
title={Reduced-Reference Quality Assessment Based on the Entropy of DWT Coefficients of Locally Weighted Gradient Magnitudes},
author={Golestaneh, SeyedAlireza and Karam, Lina J},
journal={IEEE Transactions on Image Processing},
volume={25},
number={11},
pages={5293--5303},
year={2016},
publisher={IEEE}
}
Some example results of distorted image along with its locally weighted gradient maps (e.g. Child)
The scatter plots of the prediction results of REDLOG versus the subjective score in database of LIVE, CSIQ, TID 2013, TID2008, QualTex, and TOYOMA. Each point stands for a distorted image in the database.
Examples of reference and distorted images for different distortions along with their locally weighted gradient maps:
