Vision and modeling of specular surfaces

Abstract

We present the combination of a sensor based on �Phase-Measuring Deflectometry�and a new numerical algorithm to obtain the shape of specular free-formsurfaces. The sensor measures the local slope of the surface whichthen is used to reconstruct the object�s shape. The sensor is calibratedand yields absolute slope data. We solved the inherent ambiguityof deflectometric sensors using a novel stereo method for specularsurfaces. The presented reconstruction algorithm interpolates theslope data and thus yields an analytic representation of the surface.It can deal with highly curved objects, is stable against noise,and does not require equidistantly sampled data. With the combinationof the sensor and the robust numeric algorithm we are able to obtainthe global shape of specular free-form surfaces with micron accuracy.Local details can even be measured in a range of a few nanometersdepth variation.