Arthur Leclaire


A Texton for Fast and Flexible Gaussian Texture Synthesis

Gaussian textures can be easily simulated by convolving an image sample with a conveniently normalized white noise. However, this procedure is not very flexible (it does not allow for non-uniform grids in particular), and becomes computationally heavy for very large domains. A Gaussian texture can also be approximated by a high-intensity discrete spot noise (DSN), obtained by summing randomly-shifted copies of a kernel along the points of a Poisson process.

In this talk, we propose an algorithm that summarizes a texture sample into a synthesis-oriented texton, that is, a small image for which the DSN simulation is more efficient than the classical convolution algorithm. Using this synthesis-oriented texture summary, Gaussian textures can be generated on-demand in a faster, simpler, and more flexible way.

Slides [pdf]

Related material

Project Page (Synthesis-Oriented Texton)

The Speaker

Arthur Leclaire is finishing his PhD Thesis and working as a young teacher and researcher at Université Paris-Descartes and ENS Cachan.