Underwater imaging plays a key role in marine observation and exploration, supporting various applications such as object surveillance, geographic surveys, resource exploration, and marine ecosystem protection. However, because of the complex nature of the underwater environment, raw captured images are degraded by physical processes such as light absorption and scattering, which results in color deviation, low contrast, and uneven illumination. When a light ray propagates under water, i.e., within a scattering medium it is attenuated and scattered as a function of wavelength and distance traveled — hence it is subject to radiometric distortions. That means that a given ray loses radiance along its path but gains back-scattered non-informative signal. The direct signal, carrying information, is furthermore forward-scattered, which induces a blur-like effect. Undoing this effect in a physically-based manner falls in the realm of underwater image-enhancement or image restoration. Due to the limited communication bandwidth and high latency of the underwater environment, on-board processing is mandatory. It reduces reliance on high-latency acoustic communication links; however, it also imposes more constraints on computational power, memory, and energy consumption.
The Lab-STICC, with it affiliation to the INS2I institute of the CNRS, is a research unit historically recognized in Brittany and in France in the field of ICTS. It has a proven capacity to cover a broad scientific spectrum around digital sciences, and in particular with this ability to address various disciplinary fields (Information Theory, Waves & Materials, Embedded Electronics and Computing, Data Sciences, Communication and Signal Detection, Human-Machine Interfaces,…) following multiple themes/application sectors: maritime environment, communicating objects, defense, space, health, security, and robotics. The SHAKER team is specialized in optimization of software / hardware systems according to the constraints and hazards related to their environment. Its work centers on the co-design of hardware and software architectures, the development of dynamically adaptable systems that can respond to both environmental conditions and application needs, and the verification of system behavior, including temporal properties. The team also develops practical tools to support design, simulation, and execution tracing.
Joint supervision – the students will be supervised by two teachers/researchers from the University of Brest and the University of Kiel.
In this internship, the objective is to perform a comprehensive survey and analysis of state-of-the-art techniques on underwater image enhancement/restoration. The algorithms are evaluated and compared using two primary criteria: (1) output image quality, (2) real-time performance. 1. Image quality compared to ground-truth: In terms of novel views, i.e. photometric consistency: PNSR, SSIM, LPIPS in terms of radiometric ground truth, Reproduction Angular Error (RAE) or even 2-Norm. 2. Real-time performance is assessed using metrics: computation time, memory usage, and energy consumption.
Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Education and Culture Executive Agency (EACEA). Neither the European Union nor EACEA can be held responsible for them.
