Predicting the elastic properties of carbonate rocks is crucial for the oil industry. However, the standard models that estimate effective elastic properties in porous media have many limitations in carbonate rocks. One main reason is highly complex pore-space structures, produced in carbonates by diagenesis and other geological processes, which create heterogeneities at several scales. Recently developed image acquisition systems, based on X-ray computed tomography, allow description of the spatial distribution of grains and pores in scanned samples at high resolution. Numerical simulations can then predict the elastic properties using the geometry of each phase (grains matrix and pore space). In this paper, we apply a new efficient automatic segmentation technique based on bi-level thresholding to separate grains and pores phases in 3D X-ray CT scans. Then we assess the ability of a finite-element simulation technique commonly used on sandstones to estimate the elastic properties of carbonate rocks under different fluid-saturation conditions.