Curve and surface models to drive 3D reconstruction using stereo and shading

Abstract

We propose a reconstruction technique which consists in defining a parametric surface lying on a closed curve by setting the extremal behavior of the surface along the curve on which it lies. Therefore, we most build a set of curves in space from a stereo reconstruction of matched and closed contours in an image pair. Then, we use photometric information extracted near the image contours to define the local behavior of the surfaces lying on the closed curves. In order to build such a surface we use the duality existing between a stereo-based reconstruction which is able to determine the position of points in the scene space and photoclinometry which gives information about the shape of the objects in the scene. Our geometric model is composed of a radial set of triparametric Gregory patches for each surface to reconstruct. The topological concepts introduced by these kind of surfaces and the use of photometric models gave a framework for both image analysis processing and reconstruction geometrical processing.