AUTOMATED IMAGE-ANALYSIS OF ICE THIN-SECTIONS - INSTRUMENTATION, METHODS AND EXTRACTION OF STEREOLOGICAL AND TEXTURAL PARAMETERS

Image-analysis procedures have been developed for a simple PC-based sy stem to evaluate the textural features of ice thin sections automatica lly. The method yields parameters that describe both ice crystals and pores in an objective, reproducible manner. Specimens are recorded in linearly and circularly polarized light (ice grains) as well as in tra nsmitted and incident plain light (pores) under standardized condition s. After preparative filtering, images are segmented through union of two Sobel-filtered images (ice crystals/grain boundaries) or through t hresholding (ice crystals/pores). The quantification of texture is bas ed on the evaluation of image contrast as well as on linear analysis a nd non-linear transforms to obtain information on grain-sizes and thei r distribution. In the study of ice pores, stereological effects of fi nite section thickness have to be corrected. The discussion focuses on errors involved in automated textural analysis and the glaciological yield of measured and derived parameters.