•A novel approach to generate functionally graded cellular structures was evaluated.•The relationships between the method parameters and cell size were investigated.•Variation in cell edge length was greatest for the Voronoi connection scheme.•There was approximately an 8% underestimation of cell area for both target regions.•Practical implementation implications were discussed for tackling 3D problems. The spatial variation of cell size in a functionally graded cellular structure is achieved using error diffusion to convert a continuous tone image into binary form. Effects of two control parameters, greyscale value and resolution on the resulting cell size measures were investigated. Variation in cell edge length was greatest for the Voronoi connection scheme, particularly at certain parameter combinations. Relationships between these parameters and cell size were identified and applied to an example, where the target was to control the minimum and maximum cell size. In both cases there was an 8% underestimation of cell area for target regions.