The paper is concerned with the development and application of an automatic algorithm of image analysis for detection of intercrystalline microcracks in dolomite structures in the laboratory conditions. In sections of dolomite rock samples were duly prepared. For each measurement field on thin sections we considered a set of input color images obtained using an optical polarizing microscope with different polarization set-ups. Then color system transformation was performed, followed by rock gram and intercrystalline microcracks segmentation for each color image. The results confirm the adequacy of the applied method of image analysis in ale segmentation of the microscope images of intercrystalline microcracks in dolomites. The newly-developed algorithm may facilitate the petrographical and stereological studies of rock structures.


A description and analysis of rock structures in microscale (e.g. under a polarization microscope) is very important from the standpoint of geo-scientists.

Computer image analysis techniques are vital for rock structure analysis. The paper describes results achieved by applying the methods of image analysis of mathematical morphology to geometrical analysis of rock intercrystalline microcracks.

Techniques used for detection and analysis of Mirocracks in rock structures include: stereology Mlynarczuk 2000), image analysis (Bodziony et al. 1993, Obara & Mlynarczuk 2004), acoustic emission (Maji et at. 1990), computer tomography (Durham & Beiriger 1985), computer modelling (Shao 1999).

The paper summarises the results of research on agonthms for automatic segmentation of all Intercrystalline microcracks in a dolomite structure. This method can vastly facilitate the rock structure analysis.

Input images

Thin sections from dolomite rock samples are prepared And then observed under a polarization microscope with a CCD camera connected via a video card to a Pc. field se thin sections several random measurement are t s:re selected and for each of these 7 color images are taken: 6 images with two crossed nicols (each of them at a different cross angles) and one with one nicol (Fig.1a, b)

(Figure in full paper)

Each of those images had the 688 × 566 pixels resolution (related to the size 1420 × 1065 μm)and 16,7 millions of colors in the RGB color system. Grains segmentation method These input images were transformed from the RGB system into CIELab color system (Fig. 2a, b, c) (Wyszecki & Stiles 1982).

The CIELab color system provides a good color discrimination (Wesolowski 1999). After color transformations we observed that the channel ‘a’ (Fig. 2b) is relatively insensitive to color intensity of segmented objects in the input images. That property allowed a significant improvement of the grain and microcracks segmentation technique.

(Figure in full paper)

Several image analysis operations are performed for each of measurement field on thin section:

  • RGB to CIELab color system transformation of 6

  • microscope images (taken with two crossed nicols), (Fig. 2a, b, c), morphological gradients of 6 ‘a’ channel images (exemplary ‘a’ channel image shows in Fig. 3a),

  • maximum of all gradient images, (Fig. 3b),

  • finding the local minima of maximum image, (Fig.3b),

  • watershed method.

This content is only available via PDF.
You can access this article if you purchase or spend a download.