Generation, migration, trapping, and discovery of petroleum, as well as its primary and enhanced recovery, depend to a large extent upon the geometry of the porous microstructure. The determination of petrophysical rock properties is difficult because of their complexity ranging from apparent random structures to configurations with heterogeneities at different scales. We present a method of visualizing a realistic 3D representation of these structures. While rock descriptions always contain many deterministic elements and features, some elements are most conveniently generated by stochastic models. The primary purpose of this study was to survey ways in which stochastic modeling might be used to create rock images, obtain realistic representation of rock textures, and provide basic model control and visualization capability. We use an "octree" data structure to represent rock models of arbitrary precision. Objects created using this technique are collections of subvolumes of variable sizes each having the same physical properties.
A reconstruction of porous media results in a pore network where the total porosity is the same as that of the actual rock. The model and techniques presented have been designed in an extensible fashion to enable the development of algorithms for a simulation of geometrical models, along with other simulated properties of the rock models such as electrical resistivity, fluid flow, permeability, and acoustic velocity. The developed methodology may be used to do: quick evaluation of rock parameters, calibration, teaching, determination of the estimation sensitivity, and testing of the computer software for image analysis.