3D surface reconstruction from multiple SEM images has been widely addressed in the scientific literature. Different approaches exist, most of them use variants of Digital Image Correlation (DIC) [1][2].

DIC methods perform well on textured areas of images but provide limited results on uniform or flat zones. Approaches using segmentation techniques, that match discernible regions of images, provide a solution to solve this problem on low-textured regions [3]. However, these methods are more computationally expensive and are less accurate than DIC on textured areas. Our objective is therefore to combine these two methods. One preliminary step is to differentiate textured areas, where DIC can be used, from low-textured areas, where segmentation techniques are needed. This step will be addressed in this paper, and we propose a new approach using morphological operators to do that.

Our approach uses the watershed operator with morphological gradient and markers obtained with h-minima [4]. On each region, average gradient value is computed. Automatic thresholding and post processing using opening by criteria [5] is then performed to differentiate textured areas from low-textured ones.

Comparison of our approach with more standard methods of texture detection such as local entropy of local variance calculations [6] will be presented.

A major benefit of our approach is that transition between textured areas and low-textured ones strongly follow real contours of flat zones, limiting interpolation problems between DIC and methods using segmentation techniques.

We will illustrate our texture detection algorithm by applying it on 3D surface reconstruction with SEM images of crystals of zeolites and catalyst with alumina supports.

[1] J.V. Sharp (et al), 1965. Automatic Map Compilation Using Digital Techniques. Photogrammetric Engineering. Vol. 31, No 2, pp. 223-239.
[2] Accurate 3D Shape and Displacement Measurement using a Scanning Electron Microscope. PhD Thesis, University of South Carolina, Institut National des Sciences Appliquées de Toulouse, Juin 2005.
[3] Jean-Charles Bricola, Michel Bilodeau, and Serge Beucher. A top-down methodology to depth map estimation controlled by morphological segmentation.
[4] Serge Beucher. Segmentation d’images et Morphologie Mathématique. PhD thesis, Ecole Nationale Supérieure des Mines de Paris, Juin 1990.
[5] Thomas Walter. Application de la Morphologie Mathématique au diagnostic de la Rétinopathie Diabétique à partir d’images couleur. PhD thesis, Ecole Nationale Supérieure des Mines de Paris, 2003.
[6] K. Itoh, A. Hayashi, and Y. Ichioka, “Digitized optical microscopy with extended depth of field,” Appl. Opt., vol. 28, no. 15, pp. 3487–3493, 1989.

is that transition between textured areas and low-textured ones strongly follow

Figures:

Fig. 1: Filtered SEM image of alumina supports.

Fig. 2: Texture detection on figure 1. Green: textured regions. Red: low-textured regions.

Fig. 3: Filtered SEM image of crystals of zeolites.

Fig. 4: Texture detection on figure 3. Green: textured regions. Red: low-textured regions.

« Back to The 16th European Microscopy Congress 2016

EMC Abstracts - https://emc-proceedings.com/abstract/a-morphological-approach-for-texture-detection-application-to-sem-stereo-reconstruction/