Local random-phase noise for procedural texturing

Abstract

a) Noise based on a continuous target spectrum. (b) Structure-preserving procedural texture deduced from an example. Figure 1: Local random-phase noise can approximate an arbitrary power spectral density. It provides high-speed procedural reproduction of Gaussian patterns defined by a continuous spectrum (a) or by an input example. A broader range of procedural textures by example can be generated by preserving input structures such as skin wrinkles in (b). AbstractLocal random-phase noise is a noise model for procedural texturing. It is defined on a regular spatial grid by local noises, which are sums of cosines with random phase. Our model is versatile thanks to separate sampling in the spatial and spectral domains. Therefore, it encompasses Gabor noise and noise by Fourier series. A stratified spectral sampling allows for a faithful yet compact and efficient reproduction of an arbitrary power spectrum. Noise by example is therefore obtained faster than state-of-the-art techniques. As a second contribution we address texture by example and generate not only Gaussian patterns but also structured features present in the input. This is achieved by fixing the phase on some part of the spectrum. Generated textures are continuous and non-repetitive. Results show unprecedented framerates and a flexible visual result: users can control with one parameter the blending between noise by example and structured texture synthesis.

On Demand Solid Texture Synthesis Using Deep 3D Networks

“…The results in Figures and prove the capability of the proposed model to synthesize photo‐realistic textures. This is an important improvement with respect to classical on demand methods based on Gabor/Local Random‐Phase (LRP)‐noise [GSV*14]. High‐resolution examples are important in order to obtain more detailed textures.…”

“…Creating realistic textures with a procedural noise is a trial and error process that necessitates technical and artistic skills. Some procedural methods alleviate this process by automatically estimating their parameters from an example image [GD95, GLLD12, GSV*14, GLM17]. However, they only deal with surface textures and most photo realistic textures are still out of the reach of these methods.…”

A multiscale model for rain rendering in real-time

“…Sparse convolution noise [12] is usually defined using a uniform Poisson distribution of kernels [13]. As recently introduced by Gilet et al [14], simpler distribution schemes, such as regular or jittered grids, can also be used in conjunction with various different kernels, thus providing an accurate control over local and global features. As our goal is to represent a controllable rainfall texture, we propose to define the Visible Streaks Region as a convolution of rain streaks kernels distributed along a simple regular grid.…”

Texton Noise

“…However, recent contributions addressing noise by example [LVLD10, GLLD12, GSV*14] demonstrated that a fair amount of textures can be produced directly using a single procedural noise. The main advantage of this new noise by example approach is to propose algorithms that automatically estimate the noise parameters instead of relying on creative non‐linear tweaks.…”

“…To ensure a fast evaluation and a compact representation of the noise, the texton is modelled as a generic bilinearly interpolated function having a small support, that is, a small GPU texture. Our main contribution is to show that this simple noise model is sufficient to reproduce any Gaussian texture, with a texture analysis and a procedural evaluation that are several orders of magnitude faster than recent competing approaches [GLLD12, GSV*14] since they do not rely on the usual spectral decomposition/frequency sampling approach. The main features of texton noise are the following (See Figure ): 1.…”