Thresholding of micro-ct images for morphological analysis of trabecular bone specimens

In this paper the authors present the analysis performed over images obtained with a micro – CT system having a resolution of about 10 micrometers. The images are relative to human cancellous bone specimens. The aim is to binarize such images in order to ascertain bone tissue areas with respect to void ones. The task is somewhat complicated because of the presence of physical and reconstruction artefacts within the said void regions, so the grey level do not drop to zero. The whole histogram is thus composed by two main components that should be separated by determining a proper grey level threshold. We perform the image binarization firstly using the Otsu algorithm taken as a reference and then with a specifically conceived method. The first results allowed to model the histogram as two Gaussian function plus another component, probably due to the transition regions between bone an void areas.