Underwater image restoration is challenging due to the pervasive and simultaneous presence of multiple degradations, including turbidity, uneven illumination, color distortion, and motion blur. Existing approaches typically address individual degradation types and therefore fail to generalize to complex real-world underwater conditions where these artifacts coexist. To overcome this limitation, we propose a novel three-stage sequential framework for progressive underwater image enhancement. First, Contrast Limited Adaptive Histogram Equalization (CLAHE) is applied to mitigate uneven illumination and enhance local contrast. The contrast-enhanced image is then processed by an adaptive color correction algorithm that estimates the dominant color channel intensity and compensates for color casts across the red, green, and blue channels. Finally, the refined image is passed to a data-driven diffusion-based restoration model trained using noise parameters derived from a unique turbidity-graded laboratory dataset. This dataset enabled the model to predict and quantify turbidity-related noise and reconstruct an enhanced underwater image. The proposed framework is evaluated on the Underwater Image Enhancement Benchmark (UIEB) and the Turbid Underwater Dataset (TUDS), where quantitative and perceptual results demonstrate consistent improvements at each stage of the pipeline. In particular, we report improved quantitative performance in terms of Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) compared with recent enhancement methods on a paired subset of the UIEB, indicating effective structural and perceptual reconstruction.
Iqbal et al. (2026) studied this question.