Generalized Rectifier Wavelet Covariance Models For Texture Synthesis

TL;DR

This paper introduces a new wavelet-based statistical model for texture synthesis that enhances visual quality and rivals state-of-the-art CNN-based models by leveraging non-linear wavelet representations.

Contribution

It proposes a novel family of wavelet-based statistics with generalized rectifier non-linearity, improving classical models and achieving high-quality texture synthesis.

Findings

Significantly improved visual quality over classical wavelet models

Achieved synthesis quality comparable to CNN-based models

Effective on both gray-scale and color textures

Abstract

State-of-the-art maximum entropy models for texture synthesis are built from statistics relying on image representations defined by convolutional neural networks (CNN). Such representations capture rich structures in texture images, outperforming wavelet-based representations in this regard. However, conversely to neural networks, wavelets offer meaningful representations, as they are known to detect structures at multiple scales (e.g. edges) in images. In this work, we propose a family of statistics built upon non-linear wavelet based representations, that can be viewed as a particular instance of a one-layer CNN, using a generalized rectifier non-linearity. These statistics significantly improve the visual quality of previous classical wavelet-based models, and allow one to produce syntheses of similar quality to state-of-the-art models, on both gray-scale and color textures.