Background
As one of the main sources for human to perceive the underwater world, the quality of underwater images (UWIs) directly determines the development of marine-related industries. However, the complex underwater environment causes severe image quality degradation (due to light absorption and scattering) and inefficient data transmission (relying on limited communication channel resources and narrow transmission bandwidth), limiting its application. Traditional solutions have drawbacks: conventional compression technologies (e.g., JPEG) are ill-suited to underwater image distortion, while Nyquist sampling-based systems increase data burden for resource-constrained platforms like AUVs. Compressive Sensing (CS) provides an effective solution by reconstructing signals from few random measurements, reducing data volume and enhancing image quality. Its combination with underwater imaging has become a research hotspot, though challenges like unstable image sparsity remain. In-depth research on underwater image CS is of great theoretical and practical value for advancing marine exploration and information acquisition efficiency.
Challenges
Difficulty in Image Sparsity Modeling:
Underwater Images are affected by water scattering and light absorption, commonly color shift, blurred details, and uneven , which significantly weaken their sparsity in traditional transform domains (e.g., wavelet, contourlet).
Dilemma in Balance of Sampling Rate and Reconstruction Quality:
Reduced sampling rates lead to the loss of effective information. The balance between insufficient data acquisition and high-quality reconstruction is the main and core contradiction in algorithm design.
Insufficient Image Prior Information:
The reconstruction of compressive sensing highly relies on image prior information. In underwater environments, water scattering destroys the spatial correlation, and uneven illumination further distorts the image structure.
Weakness Anti-Interference Ability:
In actual marine scenarios, dynamic interferences such as suspended particle movement, water flow disturbance, and sudden changes in illumination can lead to noisy sampling data and measurement deviations.