PASTA：提示词感知的退火扩散采样，实现预留区域免训练保留

View PDF HTML (experimental)

Abstract:While Diffusion models produce high-quality images, real-world visual design requires intentionally preserving clean regions for headlines, logos, or UI elements. Existing layout-aware text-to-image methods struggle to strictly preserve reserved regions, often introducing unwanted content that complicates downstream editing. Existing approaches either manipulate attention to displace content or suppress conditioning via noise-level interventions. However, these methods sacrifice visual coherence, producing flat regions or noticeable style inconsistencies between reserved and non-reserved areas. To address this limitation, we propose a Prompt-Aware Shared-Latent Training-Free Annealed Diffusion Sampling (PASTA), a training-free framework that enforces reserved-region constraints by selectively applying semantic guidance to object-allowed regions during sampling. PASTA operates on a shared latent trajectory, where full-prompt and context-prompt guidance are fused through mask-guided sampling. An annealed schedule suppresses semantic leakage during early denoising stages while gradually restoring visual consistency in later stages. Extensive experiments demonstrate that PASTA improves reserved-region fidelity while preserving overall image realism and perceptual consistency.

Comments:	15 pages, 4 figures, 2 tables. Accepted to ICONIP2026
Subjects:	Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as:	arXiv:2610.32390 [cs.CV]
	(or arXiv:2610.32390v1 [cs.CV] for this version)
	https://doi.org/10.48550/arXiv.2610.32390 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Ren Ajiki [view email]
[v1] Fri, 23 Oct 2026 09:16:22 UTC (20,455 KB)