Particle migration through hydraulic structures like earthen dams is primarily governed by the pore network of the granular filter. A well designed granular filter will not only allow the dissipation of excess pore water pressure but also retains the fines making the structure stable under hydraulic loads. The standard filter design procedures were based on the empirical criteria obtained from particle size distribution (PSD). However, the pore constrictions which are the pathways connecting the voids present in the filter medium play an important role in the retention criteria. These constrictions can be quantified using advanced numerical (probabilistic or discrete models) and experimental techniques (computed tomography) which will help in gaining new insights in the filtering mechanism. In this study, utilizing simplistic assumptions, a three-dimensional discrete element approach along with computational geometry is employed to obtain the Constriction Size Distribution (CSD) in granular media. Various factors that govern the CSD in a granular ensemble are packing, particle morphology, roughness, PSD, and boundary stresses. This study focused on understanding the effect of coefficient of gradation and relative density for Ottawa sand on the CSD. The CSD curve becomes steeper when the relative density increases and also for a well graded material due to the narrow range of constriction sizes under these conditions. Further, the results suggest that the empirical filter design criteria are comparable with the average constriction size (DC50) obtained numerically.