A new approach for specular surface reconstruction using deflectometric methods

Large specular industrial components like engine hoods and bumpers require a multitude of image acquisition configurations for optical surface inspection: as the domain of measurement tends to decrease with curvature, the surface must be inspected part by part. Surface points along the border of neighboring and slightly overlapping patches can be estimated by a monocular stereo approach followed by numerical surface reconstruction. By initial value estimation only at patch boundaries, we achieve a computational-cost and hardware-wise optimal solution for the robust deflectometric reconstruction.