The quality of stored frozen products such as foods and biomaterials generally degrades in time due to the growth of large ice crystals by recrystallization. While there is ample experimental evidence that recrystallization within such products (or model systems thereof) is often dominated by diffusion-limited Ostwald ripening, the application of Ostwald-ripening theories to predict measured recrystallization rates has only met with limited success. For a model system of polycrystalline ice within an aqueous solution of sugars, we here show recrystallization rates can be predicted on the basis of Ostwald ripening theory, provided (1) the theory accounts for the fact the solution can be nonideal, nondilute and of different density than the crystals, (2) the effect of ice-phase volume fraction on the diffusional flux of water between crystals is accurately described, and (3) all relevant material properties (involving binary Fick diffusion coefficients, the thermodynamic factor of the solution, and the surface energy of ice) are carefully estimated. To enable calculation of material properties, we derive an alternative formulation of Ostwald ripening in terms of the MaxwellStefan instead of the Fick approach to diffusion. First, this leads to a cancellation of the thermodynamic factor (a measure for the nonideality of a solution), which is a notoriously difficult property to obtain. Second, we show that MaxwellStefan diffusion coefficients can to a reasonable approximation be related to self-diffusion coefficients, which are relatively easy to measure or predict in comparison to Fick diffusion coefficients. Our approach is validated for a binary system of water and sucrose, for which we show predicted recrystallization rates of ice compare well to experimental results, with relative deviations of at most a factor of 2.