Procedural noise using sparse Gabor convolution

Abstract

Figure 1: (Left) We present a procedural noise that offers accurate spectral control. The user can interactively manipulate the power spectrum. (Middle) We apply the noise to a surface without the need for texture coordinates, and provide high-quality anisotropic filtering. Thanks to increased expressiveness of the noise, a simple colormap is enough to produce visually interesting textures. (Right) Since our surface noise does not require a texture parameterization, the surface can evolve dynamically and even sustain topology changes. AbstractNoise is an essential tool for texturing and modeling. Designing interesting textures with noise calls for accurate spectral control, since noise is best described in terms of spectral content. Texturing requires that noise can be easily mapped to a surface, while high-quality rendering requires anisotropic filtering. A noise function that is procedural and fast to evaluate offers several additional advantages. Unfortunately, no existing noise combines all of these properties. In this paper we introduce a noise based on sparse convolution and the Gabor kernel that enables all of these properties. Our noise offers accurate spectral control with intuitive parameters such as orientation, principal frequency and bandwidth. Our noise supports two-dimensional and solid noise, but we also introduce setup-free surface noise. This is a method for mapping noise onto a surface, complementary to solid noise, that maintains the appearance of the noise pattern along the object and does not require a texture parameterization. Our approach requires only a few bytes of storage, does not use discretely sampled data, and is nonperiodic. It supports anisotropy and anisotropic filtering. We demonstrate our noise using an interactive tool for noise design.

Watercolour Rendering of Portraits

“…First, there are a range of types of coherent noise that have been developed for generating procedural textures, for instance the well known Perlin noise [19] and also more recent examples such as Gabor noise [14]. Second, such noise functions can be easily controlled, with parameters for frequency, magnitude, orientation, etc.…”

Biologically-inspired characterization of sparseness in natural images

“…Mathematically, the a i are drawn from a heavy-tailed probability distribution function p a such as that found in natural images (See Figure 2-B). Such assumptions were previously used for generating procedural noise in computer vision [8], but we focus here on a definition based on a model of sparseness in images. A similar endeavor was initiated by [20] by defining a mixture of a Dirac and a Gaussian distributions, but here, we derive it from the sparse coding observed in a set of natural images.…”

Multiple kernels noise for improved procedural texturing

“…Since then, various statistical, spectral and visual aspects have been investigated in papers [2,5,8,10,11,13,15,17]. Mainly two approaches [7] have been proposed to define noise functions: grid-based techniques that use an interpolation function (for instance a polynomial function) and sparse convolution-based techniques that use a certain convolution kernel function along with a random point distribution.…”

“…More recently hardware issues have been considered. Lagae et al [8] propose to compute procedural noise at real-time rates using a convolution-based technique [10], and to control some aspects of its spectral energy distribution by using Gabor kernels.…”

“…But it is limited by the small range of textures that it can deal with, since it assumes the input texture is purely noisy and isotropic. Gilet et al [4] propose a technique based on a more advanced subdivision of spectral domain by fitting elliptical shapes related to Gabor noise [8]. A greater variety of purely noisy textures can be processed, such as anisotropic ones, but spectral control still remains limited.…”