Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

UDBE: Unsupervised Diffusion-based Brightness Enhancement in Underwater Images

Tatiana Taís Schein, Gustavo Pereira de Almeira, Stephanie Loi Brião, Rodrigo Andrade de Bem, Felipe Gomes de Oliveira, Paulo L. J. Drews-Jr

TL;DR

Underwater visibility deteriorates with depth due to attenuation and scattering, and brightness enhancement has been underexplored. The authors introduce UDBE, an unsupervised diffusion-based brightness enhancer that uses a conditional diffusion framework with brightness conditioning and pre-processing color/SNR representations to preserve input brightness while improving visibility, all without paired data. The approach combines a two-stage pipeline (pre-processing to generate H, L, C, N and a diffusion learning stage with FiLM brightness conditioning and DDIM sampling) and a multi-term loss to balance perceptual quality, structure, and color fidelity; it achieves competitive improvements on UIEB, SUIM, and RUIE as measured by PSNR, SSIM, UIQM, and UISM. This work advances underwater image enhancement by enabling brightness restoration in unpaired settings, potentially benefiting autonomous underwater exploration and analysis.

Abstract

Activities in underwater environments are paramount in several scenarios, which drives the continuous development of underwater image enhancement techniques. A major challenge in this domain is the depth at which images are captured, with increasing depth resulting in a darker environment. Most existing methods for underwater image enhancement focus on noise removal and color adjustment, with few works dedicated to brightness enhancement. This work introduces a novel unsupervised learning approach to underwater image enhancement using a diffusion model. Our method, called UDBE, is based on conditional diffusion to maintain the brightness details of the unpaired input images. The input image is combined with a color map and a Signal-Noise Relation map (SNR) to ensure stable training and prevent color distortion in the output images. The results demonstrate that our approach achieves an impressive accuracy rate in the datasets UIEB, SUIM and RUIE, well-established underwater image benchmarks. Additionally, the experiments validate the robustness of our approach, regarding the image quality metrics PSNR, SSIM, UIQM, and UISM, indicating the good performance of the brightness enhancement process. The source code is available here: https://github.com/gusanagy/UDBE.

UDBE: Unsupervised Diffusion-based Brightness Enhancement in Underwater Images

TL;DR

Underwater visibility deteriorates with depth due to attenuation and scattering, and brightness enhancement has been underexplored. The authors introduce UDBE, an unsupervised diffusion-based brightness enhancer that uses a conditional diffusion framework with brightness conditioning and pre-processing color/SNR representations to preserve input brightness while improving visibility, all without paired data. The approach combines a two-stage pipeline (pre-processing to generate H, L, C, N and a diffusion learning stage with FiLM brightness conditioning and DDIM sampling) and a multi-term loss to balance perceptual quality, structure, and color fidelity; it achieves competitive improvements on UIEB, SUIM, and RUIE as measured by PSNR, SSIM, UIQM, and UISM. This work advances underwater image enhancement by enabling brightness restoration in unpaired settings, potentially benefiting autonomous underwater exploration and analysis.

Abstract

Activities in underwater environments are paramount in several scenarios, which drives the continuous development of underwater image enhancement techniques. A major challenge in this domain is the depth at which images are captured, with increasing depth resulting in a darker environment. Most existing methods for underwater image enhancement focus on noise removal and color adjustment, with few works dedicated to brightness enhancement. This work introduces a novel unsupervised learning approach to underwater image enhancement using a diffusion model. Our method, called UDBE, is based on conditional diffusion to maintain the brightness details of the unpaired input images. The input image is combined with a color map and a Signal-Noise Relation map (SNR) to ensure stable training and prevent color distortion in the output images. The results demonstrate that our approach achieves an impressive accuracy rate in the datasets UIEB, SUIM and RUIE, well-established underwater image benchmarks. Additionally, the experiments validate the robustness of our approach, regarding the image quality metrics PSNR, SSIM, UIQM, and UISM, indicating the good performance of the brightness enhancement process. The source code is available here: https://github.com/gusanagy/UDBE.

Paper Structure

This paper contains 22 sections, 15 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Physics-based Underwater Image Enhancement
  4. Traditional Underwater Image Enhancement
  5. Deep Learning-based Underwater Image Enhancement
  6. Methodology
  7. Problem Statement
  8. Image Pre-Processing
  9. Brightness Adjustment
  10. Color Map Generation
  11. Noise Representation Generation
  12. Combining Color and Noise Representations
  13. Diffusion-based Learning Process
  14. Diffusion Model
  15. Loss Functions
  16. ...and 7 more sections

Figures (4)

  • Figure 1: Overview of the proposed methodology for brightness restoration in underwater images. The proposed approach is composed of two main steps: $i)$ a Pre-Processing stage, to extract and represent features for the learning process; and $ii)$ Diffusion-based Learning, to teach the model to compensate for the brightness degradation in underwater images.
  • Figure 2: Qualitative comparison of restored underwater images on the UIEB (1$^{st}$ row), SUIM (2$^{nd}$ row) and RUIE (3$^{rd}$ row) datasets. From left to right are presented the raw underwater images and the results of RUIDL, UDNet, UESAM, and our approach UDBE.
  • Figure :
  • Figure :