The effectiveness of global thresholding techniques in segmenting two-phase porous media

Abstract

The effectiveness of five global thresholding techniques, to accurately segment different geomaterials, was evaluated in this work. X-ray CT images-taken from two-phase pervious concrete, glass bead, and silica sand specimens-were analyzed for evaluating five chosen methods. The core algorithms for these methods were coded using a Matlab programming language and packaged into a standalone application software. Three hundred and thirty-five image slices were provided for the pervious concrete specimen and the cropped size of this specimen was approximately 68mm in diameter. The method proposed by Kapur et al. (1985) yielded the best results qualitatively and quantitatively (e=0.28) to the laboratory and Image-Pro measured void ratios of 0.26 and 0.30, respectively. Eleven image slices were analyzed for a 10mm in diameter glass bead specimen. Once again, the method proposed by Kapur et al. (1985) gave the best results with a void ratio of 0.91, as compared to the Image-Pro void ratio of 0.89. Ten image slices, with a cropped diameter of 4.48mm, were used for the analysis of the silica sand specimen and the Otsu (1979) method was the most successful image segmentation technique, yielding a void ratio of 0.85 (Image-Pro e=0.77). From the results, it can be said that, no single image segmentation technique performs well over a wide range of material and that the performance of each image segmentation technique varies depending on the type and state of the analyzed media.