CauStereo: structure from underwater flickering illumination

Abstract

Underwater, in littoral zones, natural illumination typically varies strongly temporally and spatially. Waves on the water surface refract light into the water in spatiotemporal varying intensity. The resulting underwater illumination field forms a caustic network and is known as flicker. Studies in underwater computer vision typically consider flicker to be an undesired effect. In contrast, recent studies^1-3 show that the spatiotemporally varying caustic network can be useful for stereoscopic vision, naturally leading to range mapping of the scene. In this paper, we survey these studies. Range triangulation by stereoscopic vision requires the determination of correspondence between image points in different viewpoints. This is typically a difficult problem. However, the spatiotemporal caustic pattern effectively encodes stereo correspondences. Thus, the use of this effect is termed² <i>CauStereo</i>. The temporal radiance variations due to flicker are unique to each object point. Thus, correspondence of image points per object point becomes unambiguous. A variational optimization formulation is used in practice to find the dense stereo correspondence field. This formulation helps overcome uncertain regions (e.g., due to shadows) and shortens the acquisition time. Limitations of the approach are revealed by ray-tracing simulations. The method was demonstrated by underwater field experiments.²