Fast and Interactive Editing Tools for Spatial Models

Abstract

Multiple point statistics (MPS) algorithms allow generation of random fields reproducing the spatial features of a training image (TI). Although many MPS techniques offer options to prescribe characteristics deviating from those of the TI (e.g., facies proportions), providing a TI representing the target features as well as possible is important. In this paper, methods for editing stationary images by applying a transformation—painting or warping—to the regions, similar to a representative pattern selected by the user in the image itself, are proposed. Painting simply consists in replacing image values, whereas warping consists in deforming the image grid (compression or expansion of similar regions). These tools require few parameters and are interactive: the user defines locally how the image should be modified, then the changes are propagated automatically to the entire image. Examples show the ability of the proposed methods to keep spatial features consistent within the entire edited image.