Synthetic imaging through wavy water surface with centroid evolution.

Abstract

Imaging through a wavy water surface is a challenging task, as the wavy water surface introduces anisoplanatism effects difficult to model and track. A typical recovery method is usually involving multiple-stage processing on a pre-acquired image sequence. A new progressive restoration scheme is demonstrated, it can run simultaneously with image acquisition and mitigate both distortion and blur progressively. This method extends the anisotropic evolution in lucky region fusion with a novel progressive optical flow based de-warping scheme, centroid evolution. A comparison has been made with other state-of-art techniques, the proposed method can create comparable results, even with much less frames acquired. Experiments with real through-water scenes have also proved the effectiveness of the method.