Flow through porous media occurs in everyday life, including scientific, medical, and engineering applications. Realistic pore-scale simulations of flow frequently use discrete images (pixels in two dimensions or voxels in three dimensions) of real-life samples as inputs. Today's commonly held belief is that higher-accuracy simulations require higher-resolution images, which often result in lengthy scanning and/or simulation times. Conversely, decreasing the resolution destroys the simulation accuracy when the features of the sample (e.g., pores) are unresolved. Here, we report the discovery of superstructures in discrete images, which emerge from the sample's features and discrete mesh. These superstructures—and not the original features of the sample—control flow in low-resolution simulations. Consequently, decreases in resolution change the topology (flow “pathways”) and morphology (pore “shapes”) in the discrete image of the sample. Using permeability as an example, we present a new methodology to enhance the flow simulation accuracy for both low resolution X-ray computed tomography-imaged and computer-generated samples. This methodology is based on the novel concept of “null point,” P0, and voxel-based resolution parameter, ⁠. The presented methodology improves extraction of quantitative information from discrete images. Our findings are not limited by image dimensionality, imaging technique, or simulated processes.