Stereoscopic fracture surface analysis by SEM using advanced stereo matching

Three-dimensional analysis of fracture surface by SEM (Scanning Electron Microscope) is inevitable to clarify the cause of fracture. It is effective to analyze with stereo matching. However, conventional template matching method to measure 3-dimensional profile of fracture surface usually employs fixed template size and has deficiencies such as mismatching and consuming much processing time. Therefore, in this paper, a new stereo matching algorithm is developed, which cannot only save processing time but assure accuracy. In this new template matching method, template size is varied to assure accuracy and lessen mismatching points. In order to shorten processing time greatly, when calculating the correlation coefficient of image template pixels to search the matching templates, all the pixels in templates are not employed, but it is proposed that one pixel is chosen every 2 or several pixels in the same positions of the two templates. In order to investigate the precision and the usefulness of the proposed method, two examples of reconstructing fracture surfaces are successfully shown. It is practically realized to measure three-dimensional profile of fracture surface in personal computer.