Underwater imaging is important for scientific research and technology as well as for popular activities, yet it is plagued by poor visibility conditions. We present a computer vision approach that removes degradation effects in underwater vision, under natural illumination. We analyze the physical effects of visibility degradation. The main degradation effect is backscatter (veiling light, path radiance), and it can be associated with partial polarization of light. Thus, we present an algorithm, which inverts the image formation process, thereby recovering good visibility.
The algorithm is based on a couple of images taken through a polarizer at different orientations. As a by-product, a distance map of the scene is also derived. We successfully demonstrated our approach in experiments conducted in the sea. In addition, we analyzed the noise sensitivity of the recovery. The noise is amplified as a function of the object distance. Hence, in a recent paper, we propose a regularization method that adapts to the mentioned distance map. In an additional work, we look at the problem of underwater spatiotemporal illumination patterns (flicker, caustics), which are caused by surface waves. We attenuate these patterns, creating images which appear as if taken under much more stable and uniform illumination.
Fig. C - Attempts to improve the images using unsharp masking and histogram equalization.