Underwater image
Title: Underwater Image Enhancement Based on Neural Rendering 
Dates:  2023
University of Plymouth staff: Dr Dena Bazazian (PI), Professor Kerry Howell 
Underwater Image Enhancement Based on Neural Rendering is a research project that explores the use of neural rendering and deep learning techniques to improve the quality and interpretability of underwater images. 
Underwater environments often suffer from visibility degradation caused by light absorption, scattering, colour attenuation, and water turbidity, making it difficult to capture clear and accurate visual information. 
The project aims to reconstruct visually enhanced underwater scenes that better represent the true appearance of objects and environments, supporting both human observation and automated vision systems. 
 

Project objectives

The project seeks to develop advanced neural rendering methods capable of restoring image quality and recovering lost visual details in underwater environments. Specific objectives include:
  • improving colour fidelity, contrast, and image sharpness; 
  • compensating for the effects of light scattering and absorption; 
  • generating realistic scene representations; and enhancing the performance of downstream computer vision applications such as object detection, segmentation, mapping, and autonomous underwater navigation. 
The project also aims to create scalable solutions that can be applied to a variety of marine and underwater monitoring scenarios. 

By integrating neural rendering with underwater image enhancement, this project aims to bridge the gap between degraded underwater observations and accurate scene understanding, enabling clearer insights into the submerged world.

Dena BazazianDr Dena Bazazian
Lecturer in Robotics and Machine Vision

 
Underwater imaging is essential for a wide range of applications, including marine ecosystem monitoring, offshore infrastructure inspection, underwater robotics, archaeology, and environmental research. However, the optical properties of water present significant challenges to image acquisition. As light travels through water, it is absorbed and scattered, leading to colour distortion, reduced contrast, blurred features, and limited visibility. 
These degradations become increasingly severe in deep, turbid, or dynamic underwater environments, often reducing the effectiveness of both human analysis and computer vision algorithms. As a result, obtaining reliable visual information remains a major challenge for underwater sensing and exploration. 
This project addresses these challenges by leveraging neural rendering techniques to model the complex interaction between light, water, and underwater objects. Using data-driven learning approaches, the system can infer and reconstruct the underlying scene appearance, producing enhanced images with improved colour balance, visibility, and detail preservation. 
Neural rendering enables the integration of geometric, photometric, and environmental information to generate more realistic representations of underwater scenes than traditional enhancement methods. By improving image quality and scene understanding, the project supports more accurate marine monitoring, autonomous underwater vehicle operation, infrastructure inspection, and scientific exploration, ultimately advancing the capabilities of underwater vision technologies. 

Centre for Decarbonisation and Offshore Renewable Energy 

In response to climate change imperatives, we are bringing together a critical mass of leading research and expertise from across the University of Plymouth. Through co-creation and collaboration with partners from business, government and key communities from across the globe, the Centre aims to be a beacon for the University’s whole-system transdisciplinary approach to solutions-oriented research, accelerating sustainable developments in decarbonisation and renewable energy.
Centre for Decarbonisation and Offshore Renewable Energy