Home

Awesome

PWC PWC PWC PWC PWC PWC PWC PWC PWC PWC PWC

KBNet: Kernel Basis Network for Image Restoration

Yi Zhang, Dasong Li, Xiaoyu Shi, Dailan He, Kangning Song, Xiaogang Wang, Hongwei Qin, Hongsheng Li

paper

<hr />

Abstract: How to aggregate spatial information plays an essential role in learning-based image restoration. Most existing CNN-based networks adopt static convolutional kernels to encode spatial information, which cannot aggregate spatial information adaptively. Recent transformer-based architectures achieve adaptive spatial aggregation. But they lack desirable inductive biases of convolutions and require heavy computational costs. In this paper, we propose a kernel basis attention (KBA) module, which introduces learnable kernel bases to model representative image patterns for spatial information aggregation. Different kernel bases are trained to model different local structures. At each spatial location, they are linearly and adaptively fused by predicted pixel-wise coefficients to obtain aggregation weights. Based on the KBA module, we further design a multi-axis feature fusion (MFF) block to encode and fuse channel-wise, spatial-invariant, and pixel-adaptive features for image restoration. Our model, named kernel basis network (KBNet), achieves state-of-the-art performances on more than ten benchmarks over image denoising, deraining, and deblurring tasks while requiring less computational cost than previous SOTA methods.

<hr />

Network Architecture

<img src = "figs/overview.jpg">

Installation

git clone https://github.com/zhangyi-3/KBNet.git
cd KBNet

pip install -r requirements.txt

# install basicsr
python setup.py develop --no_cuda_ext

Evaluation

Image Restoration Tasks

TaskDatasetTest InstructionsVisualization Results
Gaussian DenoisingUrban / CBSD / Kodaklinkonedrive
Real Image DenoisingSIDD / SenseNoiselinkonedrive
Image DeblurringDPPDlinkonedrive
Image DerainingTest1200 / Test2800linkonedrive

Citation

If you use KBNet, please consider citing:

@article{Zhang2023kbnet,
  title={KBNet: Kernel Basis Network for Image Restoration},
  author={Yi Zhang and Dasong Li and Xiaoyu Shi and Dailan He 
            and Kangning Song and Xiaogang Wang and Honwei Qin and Hongsheng Li},
  year={2023},
  journal={arXiv preprint arXiv:2303.02881},
}

Contact

Should you have any question, please contact zhangyi@link.cuhk.edu.hk

Acknowledgment: BasicSR, NAFNet, Restormer.