BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Gaussian models for fast synthesis and inpainting of microtextures
  - Arthur Leclaire (ENS Cachan)
DTSTART:20151201T150000Z
DTEND:20151201T160000Z
UID:TALK62567@talks.cam.ac.uk
CONTACT:Carola-Bibiane Schoenlieb
DESCRIPTION:Among the various existing models of textures\, Gaussian textu
 res form an \ninteresting class\, in particular because they rely on a mat
 hematical\nmodel that is very well adapted to theoretical investigations.\
 nThey allow for by-example texture synthesis and also texture mixing.\nHow
 ever\, the classical spectral simulation of Gaussian textures is not very 
 flexible (not parallel) and becomes computationally heavy for very large d
 omains.\n\nA Gaussian texture can be approximated by a high-intensity disc
 rete spot noise (DSN)\,\nobtained by summing randomly-shifted copies of a 
 kernel along the points of a Poisson process.\nThe direct simulation of th
 e DSN is simple and allows parallel local evaluation using standard comput
 er graphics techniques for the Poisson process simulation.\nStill\, the DS
 N approximation of a Gaussian texture is satisfying only for sufficiently 
 high intensity\, so that the DSN simulation is generally not faster than t
 he spectral simulation. \n\nIn our paper [1]\, we proposed an algorithm th
 at summarizes a texture sample into\na "synthesis-oriented texton"\, that 
 is\, a kernel with prescribed small support for which the \nDSN simulation
  is more efficient than the classical convolution algorithm.\nUsing this s
 ynthesis-oriented texton\, Gaussian textures can \nbe generated on-demand 
 in a faster\, simpler\, and more flexible way.\n\nIn the first part of thi
 s talk\, after describing the discrete spot noise\, we will explain how to
  compute a synthesis-oriented texton\, and show that it allows to synthesi
 ze Gaussian textures with a very low computational cost.\nIn the second pa
 rt\, we will show how Gaussian texture models can be used to address micro
 texture inpainting. We will see that in the Gaussian case\, we can rely on
  a perfect conditional simulation algorithm based on kriging estimation.\n
 \n[1] "A Texton for Fast and Flexible Gaussian Texture Synthesis" (Bruno G
 alerne\, Arthur Leclaire\, Lionel Moisan)\, proceedings of the European Si
 gnal Processing Conference (Eusipco)\, 2014 
LOCATION:MR 14\, CMS
END:VEVENT
END:VCALENDAR