岩石組織の画像解析

その自動化における現状と問題点

抄録

Textures of rocks as well as their chemical compositions have quantitative information on their formation histories. Therefore, it is important to develop common and automatic techniques of the texture description. In this paper, applications of image analyses on the quantitative representation of rock textures were reviewed. Image analyses of rock textures consist of acquisition, processing, and quantitative analyses of images. BSE image has been predominantly used for these procedures, although other kinds of images (e.g. color images acquired by image scanners and microscopes) have been used in recent years. In image processing, binary images are produced by thresholding grey scale or color images, and grains (or grain boundaries) are recognized. This procedure has a crucial role in the following image processing and quantification. Quantitative analyses of mode, grain size distribution, grain shape and other factors of rock textures are carried out based on these binary images. Many algorithms for such quantitative analyses have been developed. In two workshops for image analysis of rock textures organized in 1996 and 1997, several technical problems have been discussed.For example, difficulty in recognizing grain boundaries in rocks is a common basic problem in using any imaging tools. We should also develop new imaging tools such as X-ray CT and NMR imaging to obtain directly 3D data.