Exploring Position Encoding in Diffusion U-Net for Training-free High-resolution Image Generation
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916651374477312 |
|---|---|
| author | Zhou, Feng Cao, Pu Ma, Yiyang Yang, Lu Yin, Jianqin |
| author_facet | Zhou, Feng Cao, Pu Ma, Yiyang Yang, Lu Yin, Jianqin |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_09830 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Exploring Position Encoding in Diffusion U-Net for Training-free High-resolution Image Generation Zhou, Feng Cao, Pu Ma, Yiyang Yang, Lu Yin, Jianqin Computer Vision and Pattern Recognition 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. |
| title | Exploring Position Encoding in Diffusion U-Net for Training-free High-resolution Image Generation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2503.09830 |