# cszn/DnCNN

Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising (TIP, 2017)

Repository: https://github.com/cszn/DnCNN
Canonical: https://ross.abutalabs.com/products/dncnn
Homepage: https://cszn.github.io/
Language: MATLAB
License Family: other
Topics: image-denoising, residual-learning, super-resolution, jpeg-deblocking, matconvnet, pytorch, keras-tensorflow
Last push: 2021-10-09T22:04:36+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3673, "days_push": 1789, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1727, forks 554 (observed 2026-08-28T04:05:28.256823+00:00)

## What it is
DnCNN is the official implementation of the TIP 2017 paper 'Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising', providing deep residual CNN models for image denoising and related restoration tasks. It includes MatConvNet training/testing code with PyTorch and Keras/TensorFlow ports, and points to the newer PyTorch code in the KAIR repository.

## Use cases
- remove gaussian noise from images with a deep cnn
- denoise images using residual learning networks
- jpeg deblocking artifact removal
- super-resolution of noisy images
- train a dncnn denoising model in pytorch
- use denoiser as plug-and-play prior for image restoration
- reproduce dncnn paper results

## When to choose
- you need a well-known, heavily cited baseline for image denoising research
- you want pretrained models for gaussian denoising, srgb denoising, or jpeg deblocking
- you are implementing plug-and-play image restoration with a denoiser prior
- you want to study residual learning for low-level vision

## When to avoid
- you need actively maintained production code - the author recommends the KAIR/DPIR repositories instead
- you need a license - the repository has no license, limiting reuse
- you want state-of-the-art denoisers - newer models like DRUNet, SwinIR, or SCUNet outperform it
- you need a simple drop-in library API rather than research scripts

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing, computer-vision
- domain: computer-vision, image-processing, deep-learning, machine-learning
- platform: python
- tags: image-denoising, residual-learning, cnn, super-resolution, jpeg-deblocking, matconvnet, pytorch, keras, tensorflow, image-restoration, research-code, matlab, gpu

## Member repositories
- cszn/DnCNN (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:28.256823+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:32:16.097119+00:00, confidence not recorded.
  - readme: https://github.com/cszn/DnCNN (fetched 2026-08-28T04:05:28.256823+00:00, sha 32f1a1ca26aa)
  - homepage: https://cszn.github.io/ (fetched 2026-08-29T11:08:56.389551+00:00, sha 92110304d1e2)
- Data as of 2026-08-30T08:39:29.467469+00:00.