Learning a Maximum Entropy Model for Visual Textures using Diffusion

2026-06-15

AI Summary

Existing texture generation methods rely on handcrafted or pre-trained high-dimensional statistical features designed for other tasks, lacking an unsupervised learning mechanism tailored specifically for texture modeling. This work proposes the first unsupervised framework based on diffusion models that automatically learns a compact set of low-dimensional statistical features—only 512 dimensions—to constrain a maximum entropy probabilistic model for synthesizing high-quality visual textures. The proposed method achieves generation quality comparable to or surpassing state-of-the-art approaches, which typically employ features of approximately 177,000 dimensions, while significantly improving model compactness. Furthermore, it enables smooth, continuous synthesis of homogeneous textures through linear interpolation in the learned representation space.

Abstract

Visual textures -- spatially homogeneous image regions containing repeated elements (e.g. a field of grass, the bark of a tree) -- are ubiquitous in visual scenes and provide important cues for recognizing and analyzing materials and objects. A number of existing texture models extract essential statistics from a single texture image, and can then generate high-quality samples that are visually similar to the original by matching these statistics. However, their statistics are either hand-designed or based on a network pretrained for another purpose (e.g., object recognition). Here, we develop the first principled method for unsupervised learning of a set of statistics that are used to constrain a maximum entropy probability model. We leverage methods developed for generative diffusion models to derive training and sampling procedures, and compare these to the traditional method of sampling via matching the statistics. Despite the compactness of our trained model (512 statistics), it generates texture images whose quality is as good as or better than the current state-of-the-art model (~177k statistics). A more direct comparison of the two models, obtained by synthesizing images that are indistinguishable for one model but maximally different for the other, reveals their relative strengths and weaknesses. Finally, we show that unlike previous statistical texture models, a straight trajectory in the representation space of our model generates homogeneous texture samples that interpolate smoothly between the features of the two end points.

Problem

• visual textures
• maximum entropy model
• unsupervised learning
• statistical modeling
• texture synthesis

Innovation

• maximum entropy
• diffusion models
• unsupervised learning
• texture synthesis
• statistical modeling