Preprint
Sep 2026
Multi-History-Step SDE Inversion for Image Editing with Superior Regional Awareness
MIEdit introduces a predictor-corrector multi-history-step scheme to achieve superior editing quality with fewer steps, and mitigate heterogeneity and conflict between the multi-conditioned noise residuals and gradient terms during sampling, improving stability and editing plasticity under large edits.
Hai-Yan Wei, Yun-Long Wang, Huai-Bo Huang et al.
· 0 citations