An integrated solution for underwater image restoration in artificially lit turbid water
Tianchi Zhang, Yusong Li, Xing Liu, Mingjun Zhang · Journal of Ocean Engineering and Science · 2026
This study addresses underwater image restoration in artificially lit low to medium turbid water for vision-based AUV target detection. We propose an integrated restoration framework founded on three key innovations: a bright channel-based method for robust depth estimation, a depth-aware approach for consistent background light estimation, and a background feature-based technique for transmittance estimation under unknown attenuation conditions. Compared to existing methods, the unified framework effectively mitigates depth-related errors, reduces color distortion, and operates without prior knowledge of optical parameters in turbid water. The proposed method is validated using both a private dataset and public benchmarks. Experimental results confirm that it significantly enhances image quality under challenging illumination, demonstrating strong practical value for underwater visual applications.