3D shape reconstruction from autostereograms and stereo

We study the problem of shape reconstruction from stereo images based on a weighted area minimization process of a depth function. As a simple example we present an efficient shape reconstruction from computer generated autostereograms. A minimal surface area based correlation is applied to accurately reconstruct the surface structure embedded first in one autostereogram image and next in two or more stereo images. The minimal area approach proved itself as a useful geometric measure in recent reconstruction and enhancement applications in computer vision and image processing. Here we develop a simplified version for the O. Faugeras and R. Keriven (1998, IEEE Trans. Image Process. 7, 336-344) stereo reconstruction model and apply a weighted area measure as part of a solution to the correspondence extraction in the shape from stereo and the shape from autostereogram problems. The proposed schemes are computationally efficient and yield accurate 3D reconstructions for smooth as well as nonsmooth surfaces. (C) 2002 Elsevier Science (USA).