Diffusion Image Editing via Asynchronous Token Decoding
2026-08-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors tackle the problem of changing parts of an image based on text instructions without messing up other parts like background or layout. They propose a method called ATDEdit that edits only certain parts of the image by deciding which tokens (pieces of the image data) to change step-by-step instead of all at once. This helps keep unedited areas stable while making the desired changes. Their approach does not need extra masks or retraining and performs well in keeping images looking similar to the original while editing.
diffusion modelsimage editingtext-guided editingtoken decodingdenoisingconditional surprisallatent spacekey/value memoryPSNRLPIPS
Authors
Yang Shi, Liangsi Lu, Minzhe Guo, Yifeng Xie, Yanhui Chen, Jingchao Wang, Xuhang Chen
Abstract
Text-guided diffusion image editing aims to modify semantic attributes of an image while preserving its identity, layout, and background. However, naïvely switching the text condition during sampling often causes global drift, as denoising dynamics propagate changes across tokens and can disrupt unedited regions. To address this issue, we propose \textbf{A}synchronous \textbf{T}oken \textbf{D}ecoding \textbf{Edit} (ATDEdit), an inference-time framework that views each sampler step as a parallel update of a globally coupled token matrix and enables token-indexed condition switching with differentiated update policies. Instead of applying synchronous target-conditioned updates to all tokens, ATDEdit estimates editable locations using token-wise conditional surprisal and applies target-conditioned corrections to the selected token set. It supplies source key/value memory at keep-token positions and projects selected keep-token latent rows back to their source values; these operations promote background preservation but do not constitute a pixel-level invariance guarantee. This approach combines local editing and background preservation without external or user-provided spatial masks and without model fine-tuning. On PIE-Bench, ATDEdit achieves the strongest reported preservation metrics, including 27.44~dB PSNR and 0.055 LPIPS, while retaining competitive semantic alignment.