Recovery of 3D depth map from image shading for underwater applications

We study the problem of exploiting image shading for the recovery of an object´s 3D shape underwater. This requires the employment of image models that take into account shading effects due to the medium optical properties, as well as the surface shape and reflectance properties. A simplified image model that incorporates the attenuation of the incident light due to beam spreading, as well as medium absorption and forward-scattering of the incident and reflected light, have been employed to recover the orientation of a planar surface from image shading [Yu and Negahdaripour 1991]. We investigate the application of this model for the recovery of the 3D shape of curved surfaces. To do this, we have implemented two methods based on the generalization of the shape from shading method of Tsai and Shah [1994]. One iterative method involves recursive adjustment of the underwater image, based on the estimate of the target´s 3D shape, to correct for the shading effects due to illumination attenuation. Convergence is reached when the recovered shape, after image shading correction for the medium effects, is consistent with the object´s reflectance map. A second method estimates the object´s 3D shape directly from the underwater image shading model. Convergence properties and the accuracy of the solutions from the two methods are demonstrated using synthetic data. Results from experiments with the images of an underwater scene are presented to show performance with real data