A Comprehensive Benchmark of Underwater Vision Enhancement Techniques for Sustainable Water Management Applications
Rahul Jain, Geeta Chhabra GandhiIntroduction:
Sustainable management of marine and freshwater environments requires effective monitoring. However, in underwater visual data acquisition, the effects of light attenuation and scattering vary with wavelength, degrading image quality.
Methods:
In this study, the latest underwater image enhancement algorithms were systematically evaluated on 5 publicly available databases (UIEB, Sea-Thru, RUIE, SUIM and URPC2020). Altogether, 3 main method categories (classical processing, physical model-based approaches and deep learning methods) are compared in terms of the following objective performance metrics: mean Average Precision (mAP@0.5), Underwater Image Quality Measure (UIQM), Underwater Color Image Quality Evaluation (UCIQE), processing time and memory consumption.
results:
This benchmark is an entire indication of how image enhancement technologies are critical towards supporting effective visual surveillance systems in water management applications. Based on our analysis, the picture is more complicated with a variety of methodological approaches having a different performance specific to various management objectives and operational limitations.
Results:
Deep learning methods, especially the GAN-based approach with FUnIE-GAN, yield the highest object detection accuracy (up to 45.6% mAP@0.5). Classical methods (such as CLAHE) have the lowest computational time (12 ms) and memory usage (15 MB), making them well suited for real-time monitoring. Physical model-based approaches offer intermediate performance and a stronger theoretical basis.
Discussion:
The study highlights key method-specific compromises between enhancement quality, computational cost, and deployment practicality. In cross-dataset evaluations, it was found that results depend strongly on water conditions, highlighting the importance of selecting techniques tailored to them.
Conclusion:
Application-specific selection guidelines are provided, matching enhancement techniques to water management goals, such as biological monitoring, infrastructure inspection, pollution detection, and habitat assessment. The results contribute to SDG 14 (Life Below Water) and SDG 13 (Climate Action) by developing tools for reliable visual monitoring of aquatic ecosystems.