Semi-automatic grain border detection based on active contours for marble thinsection image processing

OData support
Dr. Csorba Kristóf
Department of Automation and Applied Informatics

In geology, thin section images are used widely for material examination. On thin section images grains can be seen which carry valuable information for the geologists. In the case of marbles manual approaches are still commonly used for this kind of examination due to disturbing elements on the images (like twin crystals, dark spots), which make any automatic image segmentation difficult. GrainAutLine - a project of the Department of Automation and Applied Informatics - aims to make the thin section processing automatic for marbles.

The goal of this thesis is to find an active contour based solution for the segmentation of marble thin section images witihin the GrainAutLine project. The solution presented in this thesis is specifically based on parametric active contours, and follows a semi-automatic approach: first the user marks small areas within each of the grains, which will serve as starting locations for the active contours. These active contours then start to progress towards the grain boundaries. A balloon force has an important role during this progress, and a special force is also introduced to forbid the contours to overlap with each other. For greater accuracy the number of points in an active contour is dynamically changed during the iteration. A special problem of marble thin section image processing is the presence of small, dark spots on the images: these spots can cause a serious deformation of the active contour in a way where loops may appear along the contour. In this thesis I show the design and development of an active contour based solution.


Please sign in to download the files of this thesis.