Scene-aware atmospheric light estimation and superpixel transmission model for underwater image restoration

Underwater image restoration remains challenging due to light absorption and scattering, which cause significant blurriness and color degradation. This study presents a tailored underwater image restoration framework, Scene-Aware Atmospheric Light and Superpixel Transmission Model (SALST), designed for scenarios with pronounced scattering effects. The framework comprises five stages: scene classification, superpixel-based transmission estimation, scene-aware atmospheric light estimation, image restoration, and adaptive color correction. A key innovation is the superpixel-based transmission model, which independently estimates transmission for each color channel, addressing limitations of traditional uniform transmission methods. In addition, the scene-aware atmospheric light estimation method employs scene-specific strategies, using global approaches for the far-field and local ones for the near-field, improving estimation accuracy. Furthermore, this study proposes a novel entropy-based scene depth estimation algorithm. Experimental evaluations demonstrate the superior performance of the proposed SALST framework compared to state-of-the-art methods, achieving notable improvements in depth estimation and deblurring under varying scattering conditions. • Proposed SALST: a novel framework for underwater image restoration in strongly scattering environments. • Developed a superpixel-based transmission model that independently estimates per-channel transmission. • Designed a scene-aware atmospheric light estimation method with distinct strategies for near/far-field scenes. • Introduced an entropy-based depth estimation algorithm for improved atmospheric light calculation.