Digital Mapping as a Tool for Quantification of Textures in Thin Sections
Abstract
The quantification of textural parameters in thin sections of rocks is the first step towards understanding the evolution of microstructural fabric and relating it to rock history. However, this essential stage is seriously limited by the time-consuming nature of manual measurements. Computer-aided edge detection is made difficult by the wide variety of grain birefringence and size, as well as defects such as fractures, alteration and twinning. I have successfully applied a range of new filters and segmentation tools from medical imaging science. The resulting raster image can be easily vectorized, and the outcome of this process is a digital map which can be further analyzed using GIS software. Important improvements using this method are better noise reduction, segmentation, and raster to vector conversion, for a multifaceted two-dimensional analysis. Instead of using the basic median filter for noise reduction, I performed an anisotropic diffusion that preserves the sharpness of grain boundaries. Next, detection of grain-edges was achieved by watershed segmentation, which treats an image as a height function, and grain edges as gradient descents between grains. It classifies pixels into regions analogous to catchment basins in a flooded landscape topology. The conversion from segmented raster image to a vectorial format is then performed and even if there is still some cleaning left, the time required to contour the thousands of crystals of a thin section is greatly reduced. Spatial registration and scaling of the polygons using GIS methods allow analyzing the thin section exactly as a digital map. The Matlab toolbox PolyLX (Lexa, 2003) is used to compute various statistical parameters such as shape descriptors (e.g., aspect ratio, compactness, etc.) and crystal size distribution. I can also produce various maps in order to detect any relationship between a grain attribute and its position. This technique will be applied to a range of plutonic and volcanic rocks in order to assess the influence of crystallization processes on the crystal shape distribution. The chosen rocks show well-defined typical textures that can be related to a known crystallization history. Systematic analysis will then permit identification of the signature and relative importance of mechanisms such as grain growth, pressure solution or textural equilibration in the crystallization history of more complex rocks. Lexa, O., 2003. Numerical approaches in structural and microstructural analyses. Ph.D. Thesis, Charles University, Prague.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2003
- Bibcode:
- 2003AGUFM.V11A..03B
- Keywords:
-
- 8030 Microstructures;
- 8439 Physics and chemistry of magma bodies;
- 8499 General or miscellaneous