The yield stress of a ductile 2D porous material with a heterogeneous distribution of pores is investigated by means of a numerical tool based on the Fast Fourier Transform. Random distributions of circular voids, with a uniform or variable pore size and with overlapping or clearly separated pores, are compared, for a same overall porosity, to microstructures obtained as the intersection of the previous ones with a boolean scheme with a larger characteristic size. For pure shear and even more under purely hydrostatic load, the numerical simulations show that the pore distribution has a strong effect on the yield stress, with the so-called “Hashin Composite Cylinders Assemblage” being the hardest microstructure and the connected clusters the softest, while disconnected clusters lead to intermediate responses. An image analysis tool for the statistical characterization of the pore distribution is then proposed. It is based on the distribution function of the distance function, the width of which increases with the presence of clusters. An additional analysis allows one to distinguish microstructures with connected clusters from those with disconnected ones.