Underwater Turbid Image Restoration Using Diffusion Models

Underwater images often face challenges due to turbidity caused by suspended particles, leading to hazy and distorted visuals. The lack of real-life underwater data also reduces the efficiency of trained models. This paper introduces a diffusion model-based denoising architecture to restore underwater turbid images. The method first quantifies the turbidity noise by optimizing the variance parameter using a dataset that replicates the diffusion process of forward noise addition. The trained U-Net architecture then iteratively reconstructs turbid images by implementing a reverse Markov diffusion chain process. In addition to visual enhancements, restored images are evaluated using perceptual evaluation measures such as entropy and the naturalness image quality evaluator. The results of this project will contribute significantly to underwater research by facilitating the monitoring of marine ecosystems and studying fish migration patterns.