Advanced reservoir characterization methods can yield geological models at a very fine resolution, containing 10(11)-10(18) cells, while the common reservoir simulators can only handle much lower numbers of cells due to computer hardware limitations. The process of coarsening a fine-scale model to a simulation model is known as upscaling. Predicting the accuracy of simulation results over an upscaled grid with respect to the fine grid is highly important, as it can yield the optimum upscaling process. In this paper, permeability-based and velocity-based a priori error estimation techniques are proposed by introducing image processing-based comparison methods in the context of upscaling. The performance of the introduced error estimation techniques as well as the contribution of the employed image processing method are investigated thoroughly over highly heterogeneous cases under various coarsening levels, permeability upscaling methods, boundary conditions, and mobility ratios. The results show the superior performance of the proposed image processing-based techniques compared to the existing methods in terms of predicting the accuracy of the water-flooding process over the upscaled model with respect to the original fine grid results.