Saliency-CNN-IMAGE-QUALITY-ASSESSMENT

We present a novel No-Reference Image Quality Asssessment (NR-IQA) method using deep convolutional neural networks combined with saliency map. This project is based on the deep learning library Lasagne and Theano. You can download the pretrained model via the link below. Note that the model was trained on local normalized images, please read the paper for more details. The saliency map used in our project is [1], matlab code.

Paper: PDF
Model: link

BIBTEX:
@Article{Jia2017,
author="Jia, Sen and Zhang, Yang",
title="Saliency-based deep convolutional neural network for no-reference image quality assessment",
journal="Multimedia Tools and Applications",
year="2017",
month="Aug",
day="22",
issn="1573-7721",
doi="10.1007/s11042-017-5070-6",
url="https://doi.org/10.1007/s11042-017-5070-6"
}

[1] Hae Jong Seo, and Peyman Milanfar, "Static and Space-time Visual Saliency Detection by Self-Resemblance", The Journal of Vision 9(12):15, 1-27, http://journalofvision.org/9/12/15/, doi:10.1167/9.12.15