THE APPLICATION OF MULTILAYER DIGITAL IMAGE-PROCESSING TECHNIQUES TO THE DESCRIPTION OF SOIL THIN-SECTIONS

A multilayer approach to the image analysis of soil thin sections is presented. Sixteen images from the same microscope field were digitized from photomicrographs obtained with different light polarizations combined with different colour filters. The light sources were plain transmitted light (PTL), between crossed polars (BXP), circularly polarized light (CPL) and CPL with gypsum plate. ERDAS: a digital imaging processing (DIP) system normally applied to remote sensing was used. This, together with the hardware, allowed the creation of multilayer images on which supervised and unsupervised classification procedures were applied to distinguish and quantify features such as quartz, clay coatings, matrix and pores. A combination of PTL and CPL images gave the best multilayer image. Inaccuracies in the classification procedures were corrected manually to produce a reference image which was compared with the images created by the classifications. The unsupervised classification had many limitations, however the supervised classification was very successful in differentiating the features.