New method improves stability and quality in one-step image diffusion editing
Beyond Temporal Smoothing: Spatial Energy Budgets Stabilize One-Step Diffusion Editing
Computer Vision and Pattern Recognition
Summary
Diffusion editing lets people change images using descriptive text but often causes unwanted distortions in the background or main object. The paper’s authors discovered that existing methods average changes over time, which can cause inconsistent updates across the image. They propose a new approach, BudEdit, that carefully controls how much editing energy is used and where it is applied, focusing only on relevant parts of the image. This approach improves the quality and stability of edits without needing extra training or complex inversion steps.
What this means in practice
- •For image editing software developers: Create more stable and accurate text-guided image editing tools by controlling where and how editing changes are applied.
- •For graphic design teams: Produce higher-quality image edits with less background distortion and faster processing when modifying images based on text prompts.
Authors
Shengxiao Zhou, Lei Luo, Jian Yang
Abstract
One-step text-guided diffusion editing is efficient but prone to spatially misallocated updates that distort the edited object and alter the background. Existing methods often improve stability by averaging the editing field across timesteps. We instead identify spatial energy misallocation as a distinct and measurable failure mode: across two independent noise draws, the residual field is essentially unrepeatable, making the background field unreliable for direct transport, while its total energy still sets a usable magnitude for the draw at hand. BudEdit turns that magnitude into an explicit budget and reallocates it to edit-relevant regions selected jointly by residual energy and cross-attention, controlling where editing energy is spent rather than averaging over timesteps. The resulting training-free, inversion-free editor spends the budget on transport and reuses it to scale a correction in a lower-noise gated refinement. The budgeted injection field matches its prescribed budget exactly and vanishes on the identified background support, by construction. On PIE-Bench with SD-Turbo, BudEdit outperforms ChordEdit under each method's reported default settings on all 11 evaluated metrics, including a $2.1$\,dB gain in background PSNR, $31$\% lower DINO, and $36$\% lower LPIPS, while improving all five editing-quality metrics and reporting the lowest runtime in the comparison.