This study presents an image-based framework for simulating the freezing process in water-saturated porous media at the pore scale. The proposed approach conceptualizes the irregular pore space as a system of overlapping spherical pores, within which the thermodynamically equilibrated configurations of water and ice phases can be determined using the Gibbs-Thomson equation. By integrating this concept with a pore-morphology-based modeling technique, we can simulate in-pore phase distributions directly on a 3D tomographic image of the porous material. Repeating these simulations at decreasing temperatures allows for the construction of the material’s freezing characteristic curve, which describes the relationship between temperature and the fraction of ice formed in its pore space. Our validation exercise shows that the proposed method can provide a freezing characteristic curve that is very close to the experimental measurements, even using a surrogate digital microstructure generated by the discrete element model. This demonstrates that the proposed image-based framework can serve as a reliable and practical tool for characterizing the freezing behavior of water-saturated porous media.