A new approach to the partitioning of the void space from volume images is presented, A variant of morphological thinning is used to partition the void space into a discrete and well-defined collection of individual pores, The approach leads to the identification of pores which are defined by criteria which are well established in the literature, Each pore is bounded by interfaces between the void space and solid matrix, and by planes erected where the hydraulic radius of the pore space exhibits local minima defined as pore necks, From the partitioned data various statistics are readily obtainable such as the pore-size and coordination distributions, and the distribution of pore surface area between interfaces with solid phase and those with neighboring pores, We also calculate two fractal dimensions related to the discrete pores which we suggest may represent an important method of characterizing the pore space, Although the method is illustrated by application to magnetic resonance imaging (MRI) data of a packed bed of uniform spheres, the algorithm is entirely general and may be applied to any fully sampled three-dimensional image or reconstructed porous medium.