StructDiff: A Structure-Preserving and Spatially Controllable Diffusion Model for Single-Image Generation

arXiv cs.CV / 4/15/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

StructDiff is a single-scale diffusion-based framework for single-image generation that targets preserving the source image’s structure and internal visual statistics without requiring external data.
The model uses an adaptive receptive field module to maintain both global and local distributions, helping it handle images with large rigid objects and strict spatial constraints.
StructDiff adds 3D positional encoding as a spatial prior to enable spatial control over positions, scale, and local details of generated objects, using PE-based manipulation for single-image generation.
The paper proposes an evaluation metric leveraging large language models (LLMs) to better assess single-image generation beyond traditional objective metrics and reduce reliance on costly user studies.
Experiments indicate improved performance over prior methods in structural consistency, image quality, and spatial controllability, with demonstrated usefulness for text-guided generation, editing, outpainting, and paint-to-image synthesis.

Abstract

This paper introduces StructDiff, a generative framework based on a single-scale diffusion model for single-image generation. Single-image generation aims to synthesize diverse samples with similar visual content to the source image by capturing its internal statistics, without relying on external data. However, existing methods often struggle to preserve the structural layout, especially for images with large rigid objects or strict spatial constraints. Moreover, most approaches lack spatial controllability, making it difficult to guide the structure or placement of generated content. To address these challenges, StructDiff introduces an \textit{adaptive receptive field} module to maintain both global and local distributions. Building on this foundation, StructDiff incorporates 3D positional encoding (PE) as a spatial prior, allowing flexible control over positions, scale, and local details of generated objects. To our knowledge, this spatial control capability represents the first exploration of PE-based manipulation in single-image generation. Furthermore, we propose a novel evaluation criterion for single-image generation based on large language models (LLMs). This criterion specifically addresses the limitations of existing objective metrics and the high labor costs associated with user studies. StructDiff also demonstrates broad applicability across downstream tasks, such as text-guided image generation, image editing, outpainting, and paint-to-image synthesis. Extensive experiments demonstrate that StructDiff outperforms existing methods in structural consistency, visual quality, and spatial controllability. The project page is available at https://butter-crab.github.io/StructDiff/.