Exploring Position Encoding in Diffusion U-Net for Training-free High-resolution Image Generation

2025-03-12

AI Summary

In high-resolution image generation, pre-trained U-Net architectures suffer from positional distortion in the latent space due to zero-padding in convolutions, leading to repetitive patterns and structural incoherence during denoising. This degradation stems from inconsistent positional encoding across scales. To address this, we propose Progressive Boundary Complement (PBC), a training-free method that dynamically constructs virtual image boundaries within feature maps to rectify positional information propagation and enhance cross-scale consistency. PBC integrates seamlessly into standard diffusion frameworks without modifying network architecture or requiring retraining. Experiments demonstrate that PBC significantly improves structural integrity and content richness in generated high-resolution images, outperforming existing alignment-based approaches on multi-scale synthesis tasks.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Generation

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Denoising higher-resolution latents via a pre-trained U-Net leads to repetitive and disordered image patterns. Although recent studies make efforts to improve generative quality by aligning denoising process across original and higher resolutions, the root cause of suboptimal generation is still lacking exploration. Through comprehensive analysis of position encoding in U-Net, we attribute it to inconsistent position encoding, sourced by the inadequate propagation of position information from zero-padding to latent features in convolution layers as resolution increases. To address this issue, we propose a novel training-free approach, introducing a Progressive Boundary Complement (PBC) method. This method creates dynamic virtual image boundaries inside the feature map to enhance position information propagation, enabling high-quality and rich-content high-resolution image synthesis. Extensive experiments demonstrate the superiority of our method.

Problem

• Addresses repetitive patterns in high-resolution image generation
• Explores inconsistent position encoding in U-Net denoising
• Proposes training-free method for better position information propagation

Innovation

• Progressive Boundary Complement enhances position encoding
• Dynamic virtual boundaries improve image synthesis quality
• Training-free method for high-resolution image generation