Method for estimating volumetric distance maps from 2D depth images

DETAILED DESCRIPTION The present invention approximates 3D distance maps from range images of real-world objects or from triangulated object models using single or multiple projected 2D depth images.
The invention can be used in systems where 2D depth images of real-world objects are obtained with range sensing devices, or in systems where the 2D depth images of virtual object models are obtained by calculating the projected distance from each point in the depth image to the object model.
These systems can include means for performing the projection particularly fast using polygon rendering hardware and a graphics z-buffer.
As stated above, distance maps are a discrete representation of the distance from grid points in the volume to the closest point on the object surface.
Using an interpolation function, such as tri-linear interpolation, distances at grid points can be used to estimate the closest distance from any point within the volume to the surface.
In addition, near the surface, the gradient of the distance map can be used to estimate the surface normal.
Distances to object surfaces can be used in a number of applications.
For example, the distance maps can be used to determine a path that stays a minimum distance away from objects or an obstacle-free path with minimum length for a mobile robot.
As a second example, distance maps can be used in volume rendering to detect surfaces and surface normals that are required for determining surface reflections in shaded volume rendering.
As a third example, distance maps can be used in haptics, where an electromechanical device allows the user to "feel" the surface of a virtual object by providing force feedback to resist the user's tool when the tool penetrates the surface of the virtual object.
Haptic feedback requires both the depth of penetration of the tool into the virtual object and the normal of the closest surface point to generate the resisting force vector.
The present invention generates 3D distance maps from 2D projected depth images