A DETERMINISTIC VERSION OF THE MULTIPLE POINT GEOSTATISTICS SIMULATION / RECONSTRUCTION METHOD WXTH. THE SIMULATED / RECONSTRUCTED VALUES ARE DIRECTLY TAKEN FROM THE TRAINING IMAGES WITHOUT PRIOR ESTIMATION OF THE CONDITIONAL

摘要

A multiple-point geostatistics method of simulating, reconstructing and scaling digital data sets, in particular images, but not exclusively. The method of the invention is based on a training image that is directly sampled each time a point in the simulation image is to be filled, and does not need the preliminary creation of a data structure storing conditional probabilities, as it is often the case in known multipoint simulation method. Accordingly, the method of the invention is less memory-demanding than conventional methods, and is better suited to simulations including continuous variables. Embodiments regarding the application to reconstruction of partial images and downscaling are also presented.