The study evaluated different segmentation methods for classifying pores and solids in computed microtomography images of soil aggregates. Simulated soil aggregate images were generated with varying porosity and partial volume effects to test the methods in absence of ground truth data. The Indicator Kriging method produced segmented images most similar to the ground truths across all cases and preserved pore characteristics best, as measured by misclassification error and a region non-uniformity metric. The study recommends using the region non-uniformity criterion and Indicator Kriging method for segmenting soil aggregate images.