Towards Universal Texture Synthesis by Combining Texton Broadcasting with Noise Injection in StyleGAN-2

TL;DR

This paper introduces a novel universal texture synthesis method combining texton broadcasting with Noise Injection in StyleGAN-2, enabling the generation of diverse textures with improved quality.

Contribution

It integrates a multi-scale texton broadcasting module into StyleGAN-2, enhancing its ability to synthesize a wide range of textures, from regular to stochastic, with a new high-resolution texture dataset.

Findings

Significantly better texture quality than existing methods

Effective synthesis of both regular and stochastic textures

A comprehensive high-resolution texture dataset created

Abstract

We present a new approach for universal texture synthesis by incorporating a multi-scale texton broadcasting module in the StyleGAN-2 framework. The texton broadcasting module introduces an inductive bias, enabling generation of broader range of textures, from those with regular structures to completely stochastic ones. To train and evaluate the proposed approach, we construct a comprehensive high-resolution dataset that captures the diversity of natural textures as well as stochastic variations within each perceptually uniform texture. Experimental results demonstrate that the proposed approach yields significantly better quality textures than the state of the art. The ultimate goal of this work is a comprehensive understanding of texture space.