weizhou-geek / DeepSRQ

Blind quality assessment for image superresolution using deep two-stream convolutional networks, published in Information Sciences 2020

DeepSRQ

Blind quality assessment for image superresolution using deep two-stream convolutional networks, published in Information Sciences 2020.

Download Databases

Three databases are involved in our experiments, inluding SRID, CVIU, and QADS. In this demo, we use the latest QADS as an example.

Extract Structure and Texture Information

matlab extract_structure.m

python extract_texture.py

Divide Images into Patches

matlab convert_patch.m

Read Data

Followed the formats as:

train_structure.txt: /data/zhouw/QADS/structure_patch32/img05_3_08/img05_3_08-1.bmp 0.62647

train_lbp.txt: /data/zhouw/QADS/lbp_patch32/img05_3_08/img05_3_08-1.bmp 0.62647

Training

python train_model.py

Testing

python test_model.py

Citation

You may cite it in your paper. Thanks a lot.

@article{zhou2020blind,
  title={Blind quality assessment for image superresolution using deep two-stream convolutional networks},
  author={Zhou, Wei and Jiang, Qiuping and Wang, Yuwang and Chen, Zhibo and Li, Weiping},
  journal={Information Sciences},
  year={2020},
  publisher={Elsevier}
}