Abstract: Using points or small dots to represent any image or 2D shape is the most basic discrete representation that derives from our geometric intuition. Point pattern synthesis refers to the generation of such arrangements of points from exemplars and is motivated by a variety of applications in computer graphics, from discrete texture generation to creative pattern design, through objects placement, scene creation, as well as distribution simplification. Building on recent developments in statistical analysis and synthesis, we present computational tools to learn distributions from exemplars and seamlessly recreate them over larger regions, for instance to populate virtual worlds with varied elements, from rocks to vegetation, either automatically or by painting with interactive brushes. Motivated by the interactive design of discrete textures, we are also targeting quasi real-time methods to efficiently extract statistical properties from an input pattern, and efficiently extend them while providing the possibility to edit the pattern in a user-friendly manner through intuitive and accessible tools such as image editors. Among the long list of challenging criteria for point pattern synthesis techniques, this talk covers multi-class and multi-attribute distributions handling, anisotropic and structured patterns replication, while presenting a research avenue toward a computational framework for point pattern design for applications such as data visualization and accessibility.
Bio: Pooran Memari est chercheuse CNRS au sein du LIX (Laboratoire d’Informatique de l’Ecole polytechnique) depuis décembre 2016. De 2011 à 2016, elle était affiliée à l’équipe de recherche en Informatique Graphique de Télécom ParisTech. Avant son affectation au CNRS, elle a effectué un postdoc à Caltech, après sa thèse chez Inria Sophia Antipolis en géométrie algorithmique. Elle s’intéresse à la modélisation géométrique et ses applications.