# Multi-stage image denoising with the wavelet transform

Source: [Multi-stage image denoising with the wavelet transform](<https://www.sota2.com/research/paper/68fa6d8f1e209131cea3a98b>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T00:37:22.464Z

Authors: Chunwei Tian, Menghua Zheng, Wangmeng Zuo, Bob Zhang, Yanning Zhang, David Zhang

Published: 2022-09-26T03:28:23.000Z

## Abstract

Deep convolutional neural networks (CNNs) are used for image denoising via automatically mining accurate structure information. However, most of existing CNNs depend on enlarging depth of designed networks to obtain better denoising performance, which may cause training difficulty. In this paper, we propose a multi-stage image denoising CNN with the wavelet transform (MWDCNN) via three stages, i.e., a dynamic convolutional block (DCB), two cascaded wavelet transform and enhancement blocks (WEBs) and a residual block (RB). DCB uses a dynamic convolution to dynamically adjust parameters of several convolutions for making a tradeoff between denoising performance and computational costs. WEB uses a combination of signal processing technique (i.e., wavelet transformation) and discriminative learning to suppress noise for recovering more detailed information in image denoising. To further remove redundant features, RB is used to refine obtained features for improving denoising effects and reconstruct clean images via improved residual dense architectures. Experimental results show that the proposed MWDCNN outperforms some popular denoising methods in terms of quantitative and qualitative analysis. Codes are available at https://github.com/hellloxiaotian/MWDCNN.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Color Image Denoising on CBSD68 (PSNR for sigma=15, 25, 35, 50)](<https://www.sota2.com/research/sota/color-image-denoising-on-cbsd68-psnr-for-sigma-15-25-35-50.md>): Color Image Denoising; CBSD68; PSNR (sigma=15): 34.18.

- [Color Image Denoising on Kodak24 (PSNR by Noise Level)](<https://www.sota2.com/research/sota/color-image-denoising-on-kodak24-psnr-by-noise-level.md>): Color Image Denoising; Kodak24; PSNR (sigma=15): 34.91.

- [Color Image Denoising on CBSD68 and Kodak24](<https://www.sota2.com/research/sota/color-image-denoising-on-cbsd68-and-kodak24.md>): Color Image Denoising; CBSD68 and Kodak24; Average PSNR: 31.38.

- [Image Denoising on CC Cross-Channel 65 (test)](<https://www.sota2.com/research/sota/image-denoising-on-cc-cross-channel-65-test.md>): Image Denoising; CC Cross-Channel 65 (test); PSNR (Canon 5D, ISO 3200): 36.97.

- [Grayscale Image Denoising on Set12 (PSNR by Noise Level)](<https://www.sota2.com/research/sota/grayscale-image-denoising-on-set12-psnr-by-noise-level.md>): Grayscale Image Denoising; Set12; PSNR (sigma=15): 32.6.

- [Grayscale Image Denoising on BSD68 (PSNR Evaluation at σ=15, 25, 35, 50)](<https://www.sota2.com/research/sota/grayscale-image-denoising-on-bsd68-psnr-evaluation-at-15-25-35-50.md>): Grayscale Image Denoising; BSD68; PSNR (sigma=15): 31.39.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2209.12394>)

[PDF](<https://arxiv.org/pdf/2209.12394.pdf>)
