Recovering the Images Distorted by Surface Waves Using an Efficient Multi-Stage Image Reconstruction Strategy

Abstract:

Imaging through a fluctuating water–air interface introduces severe image distortion and motion blur, limiting the effectiveness of underwater vehicles. Previous studies have leveraged spatiotemporal correlations in image sequences for reconstruction, achieving partial improvement. However, existing methods do not adequately address the coexistence of periodic and stochastic waves in natural fluid flows. They also neglect the influence of image structural information on restoration accuracy, resulting in suboptimal image quality. This paper proposes a multi-stage reconstruction strategy combining compressed sensing and non-rigid registration. Initially, compressed sensing removes global periodic distortions from the images. Subsequently, lucky patch fusion and non-rigid registration mitigate local distortions caused by random fluctuations. Finally, principal component analysis (PCA) reduces motion blur introduced by water waves. Experimental results demonstrate that the proposed method achieves higher convergence rates and superior reconstruction accuracy compared to current techniques. The approach effectively enhances underwater imaging quality, providing clearer and more stable visual information.

Keywords: Image restoration, Image enhancement, Compressed sensing, Non-rigid registration, Lucky patch fusion, Principal component analysis, Underwater imaging, Motion blur, Wave distortion, Multi-stage reconstruction.