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

Imaging through a fluctuating water-air interface will will cause image distortion and motion blur, which seriously restricts the application and development of underwater vehicle. In previous studies, utilizing the spatiotemporal correlation of image sequences for image reconstruction is a very effective solution. However, existing methods fail to account for the phenomenon of coexisting periodic and stochastic waves in natural fluid flows, and ignore the impact of image structure information on restoration accuracy, so high-quality restored images cannot be obtained. This paper proposes a multi-stage reconstruction strategy based on compressed sensing and non-rigid registration. First, compressed sensing is used to eliminate the global periodic distortion in the image. Then, lucky patch fusion and non-rigid registration are combined to eliminate the local distortion caused by random fluctuations. Finally, principal component analysis is applied to reduce image blur caused by water waves. Experiment results demonstrate that the proposed method has a superior convergence rate compared to current methods while maintaining the accuracy of image reconstruction.