Interactive Image-based Modeling of Macro-structured Textures[IEEE CG&A 99]:
Abstract: We describe a new hybrid approach, based on some morphology and image analysis principles for an efficient
"macro-structured" texture generation. These textures are characterized by sets of "small" shapes
distributed more or less randomly upon or inside objects surfaces such as mineral-like
granulates or dots. Conventional texture synthesis techniques based on oriented particles hinder
interactive user interventions, since they mainly perform automatic and uniform
distributions. These techniques also do not use analysis for automatic information extractions from
model pictures (for example digitized photos). In this paper, we apply the notions of distance measurement,
proximity and skeletons to provide an efficient interactive control of distributions, orientations,
sizes, etc. of macro-structures. Their shapes, as well as their local surface structures
are extracted from two orthogonal views of model pictures. The method improves
texture generation by merging two generally separated domains, namely analysis and modeling. Its general
formulation permits the efficient creation of various new textural features that existing methods cannot address.
Natural textured objects often provide the perception of small oriented shapes on
their surface: they present those features which,
in texture analysis and segmentation, are known as "macrostructures".
