Abstract: Underwater optical images are prone to problems such as reduced contrast, color distortion, uneven illumination and visibility, which affect the accuracy of defect recognition and size measurement. To improve the limitations of traditional background light solution algorithms in artificial light illumination scenes, established is an underwater image background light estimation method based on a quadtree hierarchical search strategy and a fusion of smoothness, on the maximum color difference and, on the maximum brightness multi-feature prior indicators; to solve the problem of inaccurate transmission map estimation in the area affected by artificial light illumination, constructed is an underwater image transmission map estimation method integrating improved dark channel prior theory and reverse saturation map theory; and then proposed is an underwater concrete defect degradation image restoration method based on physical model improvement. Combined with underwater image restoration experiments and actual engineering cases, the method is verified from two dimensions: qualitative evaluation and quantitative reference-free index calculation, and compared with typical underwater image restoration methods such as DCP, MMLE, L2UWE, and NUCE. The experimental results show that the method proposed can achieve the restoration and improvement of underwater concrete structure defect degradation images in scenes such as artificial light illumination and weak light turbidity, and effectively improve the imaging quality of underwater concrete defect images.