Sampling Gabor noise in the spatial domain

Abstract

Figure 1: Examples of a snake model with Gabor noise sampled on it. The bottom row shows the noise components and the control maps used to weight particular parameters of the noise. The left-most example shows standard Gabor noise, in the middle the frequency of the harmonic is weighted by a hat-profile, and in the right example the noise scale and the frequency are weighted by respective profiles. In all three cases the noise is evaluated in the model's uv-domain. AbstractGabor noise is a powerful technique for procedural texture generation. Contrary to other types of procedural noise, its sparse convolution aspect makes it easily controllable locally. In this paper, we demonstrate this property by explicitly introducing spatial variations. We do so by linking the sparse convolution process to the parameterization of the underlying surface. Using this approach, it is possible to provide control maps for the parameters in a natural and convenient way. In order to derive intuitive control of the resulting textures, we accomplish a small study of the influence of the parameters of the Gabor kernel with respect to the outcome and we introduce a solution where we bind values such as the frequency or the orientation of the Gabor kernel to a user-provided control map in order to produce novel visual effects.

Local spot noise for procedural surface details synthesis

“…Fig 7) texture to generate spatially varying patterns. This approach is similar to the one introduced by Charpenay et al [21]. We show on the left image the control over the rotation matrix R (varying horizontally) and local translation of each Gaussian (varying vertically).…”

“…Note that by interpolating values in parameter space before generating the final result, we achieve a smoothly varying appearance without blending artifacts. : Each parameter can be directly given by a control map (similar to [21]). Left: the green channel controls the spreading of the distribution while the blue channel drives the kernel rotation.…”

“…To alleviate this, a first solution is to use some spatial informations to influence spectral parameters. Charpenay et al [21] studied how to influence Gabor noise parameters through the use of control maps. Another solution is to automatically extract the parameters of the kernel functions through spectral analysis of an input exemplar.…”

Real‐time Terrain Enhancement with Controlled Procedural Patterns

“…The actual computation is based on underlying Gaussian fields, for which several generation algorithms exist. We exploit the capability of some sparse convolution algorithms to produce spatial variations [CSM14, LSD21].…”