Perception- and Fidelity-aware Reduced-Reference Super-Resolution Image Quality Assessment

Overview

Getting Started

Prerequisites

• Linux
• NVIDIA GPU + CUDA CuDNN
• Python 3.7

Dependencies

All dependencies for defining the environment are provided in requirements.txt.

Dataset

• We conduct experiments on three widely-used SR-IQA benchmarks, including WIND, QADS and RealSRQ.
• We split the dataset into training and testing sets with a ratio of 8:2. The training and testing lists are stored in train.txt and test.txt respectively, following the format: “SRimage_name#MOS#LRimage_name”. The dataloader (data/RealSRQ.py) can be modified to accommodate different txt file formats.

Instruction

use sh train.sh or sh test.sh to train or test the model. You can also change the options in the options/ as you like.

Acknowledgment

The codes are based on AHIQ. Thanks for their awesome works.

Citation

@article{lin2024perception,
  title={Perception-and Fidelity-aware Reduced-Reference Super-Resolution Image Quality Assessment},
  author={Lin, Xinying and Liu, Xuyang and Yang, Hong and He, Xiaohai and Chen, Honggang},
  journal={IEEE Transactions on Broadcasting},
  year={2024},
  publisher={IEEE}
}

Contact

For any question about our paper or code, please emial [email protected].