Modeling plant microstructure is of great interest to food engineers to study and explain material properties related to mass transfer and mechanical deformation. In this paper, a novel ellipse tessellation algorithm to generate a 2D geometrical model of apple tissue is presented. Ellipses were used to quantify the orientation and aspect ratio of cells on a microscopic image. The cell areas and centroids of each cell were also determined by means of a numerical procedure. These characteristic quantities were then described by means of probability density functions. The model tissue geometry was generated from the ellipses, which were truncated when neighbouring areas overlap. As a result, a virtual microstructure consisting of truncated ellipses fills up the entire space with the same number of cells as that of microscopic images and with similar area, orientation and aspect ratio distribution. Statistical analysis showed that the virtual geometry generated with this approach yields spatially equivalent geometries to that of real fruit microstructures. Compared to the more common algorithm of Voronoi tessellation, ellipse tesselation was superior for generating the microstructure of fruit tissues. The extension of the algorithm to 3D is straightforward. These representative tissues can readily be exported into a finite element environment via interfacing codes to perform in silico experiments for estimating gas and moisture diffusivities and investigating their relation with fruit microstructure. Copyright © 2006 Tech Science Press.