Stereoscopic and photometric surface reconstruction in scanning electron microscopy

The scanning electron microscope (SEM) is one of the most important devices to examine microscopic structures as it offers images of a high contrast range with a large depth of focus. Nevertheless, three-dimensional measurements, as desired in fracture mechanics, have previously not been accomplished. This work presents a system for automatic, robust and dense surface reconstruction in scanning electron microscopy combining new approaches in shape from stereo and shape from photometric stereo. The basic theoretical assumption for a known adaptive window algorithm is shown not to hold in scanning electron microscopy. A constraint derived from this observation yields a new, simplified, hence faster calculation of the adaptive window. The correlation measure itself is obtained by a new ordinal measure coefficient. Shape from photometric stereo in the SEM is formulated by relating the image formation process with conventional photography. An iterative photometric ratio reconstruction is invented based on photometric ratios of backscatter electron images. The performance of the proposed system is evaluated using ground truth data obtained by three alternative shape recovery devices. Most experiments showed relative height accuracy within the tolerances of the alternative devices. (author)