Ocean Remote Sensing Image Quality Assessment via Quad-Domain Feature Decomposition

Degradation of images in ocean remote sensing is primarily due to light scattering, chromatic aberration, and stochastic noise. As a result, the quality of the image content has been significantly compromised. Most ocean remote sensing image quality assessment methods that focus on visual content attributes have lower performance for complex underwater distortions. In view of the aforementioned problems, this article proposes a robust ocean remote sensing image quality evaluation framework based on multidomain feature fusion. Specifically, the framework systematically decomposes quality degradation into four orthogonal components. It quantifies radiance attenuation and chromatic imbalance within the physical model domain, extracts spectral distortions via frequency-domain analysis, identifies distribution anomalies through statistical regularity modeling, and evaluates structural fidelity by measuring edge and texture degradation. Based on extensive experiments across multiple benchmark databases, our approach has demonstrated higher accuracy and better robustness than other state-of-the-art methods. The performance gain remains significant under challenging conditions such as extreme low-light and high-turbidity environments.