TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Grayscale Image Denoising	BSD68 sigma15	KBNet	PSNR	31.98	# 2
Grayscale Image Denoising	BSD68 sigma25	KBNet	PSNR	29.54	# 1
Grayscale Image Denoising	BSD68 sigma50	KBNet	PSNR	26.65	# 2
Color Image Denoising	CBSD68 sigma15	KBNet	PSNR	34.41	# 2
Color Image Denoising	CBSD68 sigma25	KBNet	PSNR	31.80	# 1
Color Image Denoising	Kodak24 sigma50	KBNet	PSNR	30.04	# 1
Color Image Denoising	McMaster sigma50	KBNet	PSNR	30.27	# 1
Grayscale Image Denoising	Set12 sigma15	KBNet	PSNR	33.40	# 1
Grayscale Image Denoising	Set12 sigma25	KBNet	PSNR	31.08	# 1
Grayscale Image Denoising	Set12 sigma50	KBNet	PSNR	28.04	# 1
Image Denoising	SIDD	KBNet	PSNR (sRGB)	40.35	# 1
Image Denoising	SIDD	KBNet	SSIM (sRGB)	0.972	# 2
Single Image Deraining	Test1200	KBNet	PSNR	33.82	# 1
Single Image Deraining	Test1200	KBNet	SSIM	0.931	# 1
Single Image Deraining	Test2800	KBNet	PSNR	34.19	# 1
Single Image Deraining	Test2800	KBNet	SSIM	0.944	# 1
Color Image Denoising	urban100 sigma15	KBNet	Average PSNR	35.15	# 1
Grayscale Image Denoising	urban100 sigma15	KBNet	PSNR	33.77	# 2
Grayscale Image Denoising	Urban100 sigma25	KBNet	PSNR	31.45	# 3
Color Image Denoising	Urban100 sigma25	KBNet	PSNR	32.96	# 1
Color Image Denoising	Urban100 sigma50	KBNet	PSNR	30.04	# 2
Grayscale Image Denoising	Urban100 sigma50	KBNet	PSNR	28.33	# 2

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-bsd68-sigma25)](https://paperswithcode.com/sota/grayscale-image-denoising-on-bsd68-sigma25?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/color-image-denoising-on-cbsd68-sigma25)](https://paperswithcode.com/sota/color-image-denoising-on-cbsd68-sigma25?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/color-image-denoising-on-kodak24-sigma50)](https://paperswithcode.com/sota/color-image-denoising-on-kodak24-sigma50?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/color-image-denoising-on-mcmaster-sigma50)](https://paperswithcode.com/sota/color-image-denoising-on-mcmaster-sigma50?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-set12-sigma15)](https://paperswithcode.com/sota/grayscale-image-denoising-on-set12-sigma15?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-set12-sigma25)](https://paperswithcode.com/sota/grayscale-image-denoising-on-set12-sigma25?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-set12-sigma50)](https://paperswithcode.com/sota/grayscale-image-denoising-on-set12-sigma50?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/image-denoising-on-sidd)](https://paperswithcode.com/sota/image-denoising-on-sidd?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/single-image-deraining-on-test1200)](https://paperswithcode.com/sota/single-image-deraining-on-test1200?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/single-image-deraining-on-test2800)](https://paperswithcode.com/sota/single-image-deraining-on-test2800?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/color-image-denoising-on-urban100-sigma15-1)](https://paperswithcode.com/sota/color-image-denoising-on-urban100-sigma15-1?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/color-image-denoising-on-urban100-sigma25)](https://paperswithcode.com/sota/color-image-denoising-on-urban100-sigma25?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-bsd68-sigma15)](https://paperswithcode.com/sota/grayscale-image-denoising-on-bsd68-sigma15?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-bsd68-sigma50)](https://paperswithcode.com/sota/grayscale-image-denoising-on-bsd68-sigma50?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/color-image-denoising-on-cbsd68-sigma15)](https://paperswithcode.com/sota/color-image-denoising-on-cbsd68-sigma15?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-urban100-sigma15-1)](https://paperswithcode.com/sota/grayscale-image-denoising-on-urban100-sigma15-1?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/color-image-denoising-on-urban100-sigma50)](https://paperswithcode.com/sota/color-image-denoising-on-urban100-sigma50?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-urban100-sigma50)](https://paperswithcode.com/sota/grayscale-image-denoising-on-urban100-sigma50?p=kbnet-kernel-basis-network-for-image)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/kbnet-kernel-basis-network-for-image/grayscale-image-denoising-on-urban100-sigma25)](https://paperswithcode.com/sota/grayscale-image-denoising-on-urban100-sigma25?p=kbnet-kernel-basis-network-for-image)`

KBNet: Kernel Basis Network for Image Restoration

6 Mar 2023 · Yi Zhang, Dasong Li, Xiaoyu Shi, Dailan He, Kangning Song, Xiaogang Wang, Hongwei Qin, Hongsheng Li ·

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.

PDF Abstract

Code

Add Remove Mark official

zhangyi-3/kbnet official

Tasks

Add Remove

Color Image Denoising

Deblurring

Denoising

Grayscale Image Denoising

Image Denoising

Image Restoration

Rain Removal

Single Image Deraining

Datasets

BSD

Urban100

SIDD

Set12 CBSD68

McMaster

Synthetic Rain Datasets

Results from the Paper

Edit

Ranked #1 on Color Image Denoising on McMaster sigma50

Get a GitHub badge

Task	Dataset	Model	Metric Name	Metric Value	Global Rank	Benchmark
Grayscale Image Denoising	BSD68 sigma15	KBNet	PSNR	31.98	# 2	Compare
Grayscale Image Denoising	BSD68 sigma25	KBNet	PSNR	29.54	# 1	Compare
Grayscale Image Denoising	BSD68 sigma50	KBNet	PSNR	26.65	# 2	Compare
Color Image Denoising	CBSD68 sigma15	KBNet	PSNR	34.41	# 2	Compare
Color Image Denoising	CBSD68 sigma25	KBNet	PSNR	31.80	# 1	Compare
Color Image Denoising	Kodak24 sigma50	KBNet	PSNR	30.04	# 1	Compare
Color Image Denoising	McMaster sigma50	KBNet	PSNR	30.27	# 1	Compare
Grayscale Image Denoising	Set12 sigma15	KBNet	PSNR	33.40	# 1	Compare
Grayscale Image Denoising	Set12 sigma25	KBNet	PSNR	31.08	# 1	Compare
Grayscale Image Denoising	Set12 sigma50	KBNet	PSNR	28.04	# 1	Compare
Image Denoising	SIDD	KBNet	PSNR (sRGB)	40.35	# 1	Compare
Image Denoising	SIDD	KBNet	SSIM (sRGB)	0.972	# 2	Compare
Single Image Deraining	Test1200	KBNet	PSNR	33.82	# 1	Compare
Single Image Deraining	Test1200	KBNet	SSIM	0.931	# 1	Compare
Single Image Deraining	Test2800	KBNet	PSNR	34.19	# 1	Compare
Single Image Deraining	Test2800	KBNet	SSIM	0.944	# 1	Compare
Color Image Denoising	urban100 sigma15	KBNet	Average PSNR	35.15	# 1	Compare
Grayscale Image Denoising	urban100 sigma15	KBNet	PSNR	33.77	# 2	Compare
Grayscale Image Denoising	Urban100 sigma25	KBNet	PSNR	31.45	# 3	Compare
Color Image Denoising	Urban100 sigma25	KBNet	PSNR	32.96	# 1	Compare
Color Image Denoising	Urban100 sigma50	KBNet	PSNR	30.04	# 2	Compare
Grayscale Image Denoising	Urban100 sigma50	KBNet	PSNR	28.33	# 2	Compare

Methods

Add Remove

No methods listed for this paper. Add relevant methods here

Edit Social Preview

KBNet: Kernel Basis Network for Image Restoration

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit

Methods Edit Add Remove

Code

Add Remove Mark official

Tasks

Add Remove

Datasets

Results from the Paper

Edit

Methods

Add Remove