Procedural Descriptions of Anisotropic Noisy Textures by Example

Abstract

This short paper introduces a new approach to automate the creation of procedural anisotropic noisy textures by using an example. As for past approaches that allow one to obtain procedural descriptions of stochastic textures, it uses a sum of multi-scale noise functions and a combined spectral / histogram-based approach. The improvement, here, consists in better controlling the spectral domain by using Gabor noise functions. This allows us to extend the range of textures that can be addressed, while bringing a number of advantages compared to classical examplebased texture synthesis: extreme compactness, continuous definition over infinite space, easy extension to solid (even animated solid) textures and straight texture value computation in the fragment shader.This short paper introduces a new approach to automate the creation of procedural anisotropic noisy textures by using an example. As for past approaches that allow one to obtain procedural descriptions of stochastic textures, it uses a sum of multi-scale noise functions and a combined spectral / histogram-based approach. The improvement, here, consists in better controlling the spectral domain by using Gabor noise functions. This allows us to extend the range of textures that can be addressed, while bringing a number of advantages compared to classical examplebased texture synthesis: extreme compactness, continuous definition over infinite space, easy extension to solid (even animated solid) textures and straight texture value computation in the fragment shader.