Image analysis of backscattered electron (bse) images has often been used t
o determine the fraction of the phases in cement paste, such as porosity an
d anhydrous cement content. However, application of this technique to concr
ete requires separation of aggregate particles from the paste matrix in the
bse image. Binary segmentation of phases is usually based on their grey le
vels in the image, but the prey levels of aggregate frequently overlap thos
e of the other phases present. In this paper, an algorithm is presented for
separating out aggregate particles in the concrete image by a combination
of grey-level thresholding, filtering and binary operations. The results de
monstrate that the method is able to deal with segmentation of different ty
pes of aggregate. (C) 2001 Elsevier Science Ltd. All rights reserved.