Hybrid Training of Denoising Networks to Improve the Texture Acutance of Digital Cameras

Raphaël Achddou, Yann Gousseau, Saïd Ladjal
2/20/2024
eess.IVcs.AIcs.CV

No official code found. Best implementation: rachddou/acutance_loss (4 stars, MIT, last commit 2024-11-27).

Abstract

In order to evaluate the capacity of a camera to render textures properly, the standard practice, used by classical scoring protocols, is to compute the frequential response to a dead leaves image target, from which is built a texture acutance metric. In this work, we propose a mixed training procedure for image restoration neural networks, relying on both natural and synthetic images, that yields a strong improvement of this acutance metric without impairing fidelity terms. The feasibility of the approach is demonstrated both on the denoising of RGB images and the full development of RAW images, opening the path to a systematic improvement of the texture acutance of real imaging devices.

View on arXivView PDF

Code Implementations(1)

rachddou/acutance_lossCommunity73%
MIT

An official implementation of the SSVM'23 paper : hybrid training of denoising networks to improve the texture acutance of digital cameras

40Shell, PythonJun 21, 20231 years agoMIT
deep-learningimage-denoisingimage-processing

Similar Papers

VibrantLeaves: A principled parametric image generator for training deep restoration models

82% match1 repo52025

Reducing Texture Bias of Deep Neural Networks via Edge Enhancing Diffusion

81% match1 repo42024

Directing Mamba to Complex Textures: An Efficient Texture-Aware State Space Model for Image Restoration

81% match2025

Multi-Scale Representation Learning for Image Restoration with State-Space Model

80% match2024

Learning Distortion Invariant Representation for Image Restoration from a Causality Perspective

80% match1 repo502026

Modeling Dual-Exposure Quad-Bayer Patterns for Joint Denoising and Deblurring

80% match1 repo352024

Cite this paper

@article{achddou2024hybrid,
  title  = {Hybrid Training of Denoising Networks to Improve the Texture Acutance of Digital Cameras},
  author = {Raphaël Achddou and Yann Gousseau and Saïd Ladjal},
  year   = {2024},
  eprint = {2404.07212},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2404.07212v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2404.07212.svg)](https://codeofpaper.com/paper/2404.07212)

Discussion