Rethinking Conditional Generation for Underwater Salient Object Detection

2026-07-02

AI Summary

Underwater images suffer from low contrast, uneven illumination, and color distortion due to light scattering and absorption, severely degrading salient object detection performance. To address these challenges, this work proposes a Degradation-aware Conditional Generative Network (DCGNet), which uniquely integrates a physics-informed conditional generation mechanism with a lightweight diffusion Transformer. The framework incorporates a Dynamic Multi-Granularity (DMG) module, a Pseudo-Depth-Guided Underwater Physical Prior (UPP) module, and a Spatial Gaussian Saliency Prior (USG) to jointly alleviate boundary blurriness and background interference. Extensive experiments on five underwater benchmarks, including USOD10K, demonstrate that DCGNet significantly outperforms current state-of-the-art methods, confirming its effectiveness and robustness in complex underwater scenarios.

Abstract

Salient Object Detection in underwater images remains challenging due to low contrast, uneven illumination, and color distortion caused by scattering and absorption effects, which limit the effectiveness of conventional SOD methods in underwater environments. To address these challenges, we propose a Degradation-aware Conditional Generation Network (DCGNet), specifically designed to construct reliable conditional features for underwater saliency generation. First, we design a Dynamic Multi-Granularity module (DMG) grounded in the human visual system to robustly detect salient objects of varying scales with blurred boundaries. Then, we develop an Underwater Physics-Prior module (UPP), which utilizes pseudo-depth guidance to estimate underwater light attenuation and backscatter, thereby restoring degradation-aware RGB features and mitigating color distortion and boundary ambiguity. Based on the physics-guided representation, we introduce an Underwater Spatial Gaussian module (USG), which constructs a spatial Gaussian saliency prior from the strongest guided response to enhance object-centered salient regions and suppress cluttered underwater backgrounds. In addition, a lightweight timestep-adaptive Diffusion Transformer (DiT) bottleneck is inserted into the denoising decoder to refine fused features at different diffusion timesteps. Comprehensive experiments on USOD10K, USOD, CSOD10K, MAS3K, and RMAS demonstrate that DCGNet significantly outperforms existing state-of-the-art methods, verifying its potential for complex underwater visual applications.

Problem

• Underwater Salient Object Detection
• Low Contrast
• Color Distortion
• Uneven Illumination
• Scattering and Absorption

Innovation

• Degradation-aware Conditional Generation
• Underwater Physics Prior
• Dynamic Multi-Granularity
• Spatial Gaussian Saliency Prior
• Diffusion Transformer