This work closes the gap and eliminates the problem of availability of correct state-of-the-art local thresholding algorithms for X-ray tomography image segmentation. We provide a free library with graphical user interface that implements all major binarization algorithms—indicator kriging, region growing, converging active contours and Markov random field methods. MAZA library (MAjor binariZAtion library) also provides filters for image pre-processing. In addition, we explored and reported the sensitivity of the parameters of the segmentation methods and their computational complexity in terms of CPU time. There was not a clear winner among implemented segmentation algorithms and one needs to choose an approach depending on the X-ray image at hand and its quality.
Vasilyev et al. (Mon,) studied this question.