Edit-GRPO: A Locality-Preserving Policy Optimization Framework for Image Editing

2026-05-16

AI Summary

This work addresses the challenge in image editing where existing methods, relying on global policy updates, often compromise spatial locality and introduce distortions in non-edited regions. To mitigate this, the paper proposes a novel reinforcement learning framework that explicitly decouples editing and preservation objectives, integrating a locality-preserving mechanism directly into the policy optimization process for the first time. By leveraging region-aware allocation of optimization signals and enforcing spatial locality constraints, the approach aligns policy updates with the spatial structure of the editing task. Extensive experiments across diverse editing scenarios demonstrate that the method significantly outperforms current state-of-the-art techniques, achieving strong editing efficacy while effectively suppressing boundary inconsistencies and contextual distortions, thereby confirming its generality and robustness.

Abstract

A fundamental challenge in image editing lies in preserving spatial locality: edits should improve targeted content without inadvertently altering surrounding regions. However, most optimization-based editing approaches treat images as holistic entities, causing global policy updates that undermine locality and introduce undesired context changes. We observe that this issue stems from a mismatch between localized editing intent and globally applied optimization signals. Motivated by this insight, we propose Edit-GRPO, preserving Locality while optimizing image editing, a locality-preserving policy optimization framework that explicitly decouples editing and preservation objectives. By assigning region-specific optimization signals to edit and non-edit areas, Edit-GRPO aligns policy updates with the spatial structure of editing tasks, enabling localized improvements while maintaining global visual coherence. This design effectively suppresses common artifacts such as context distortion and boundary inconsistency. Extensive experiments across diverse image editing scenarios demonstrate that Edit-GRPO significantly improves locality preservation while maintaining strong editing performance compared to existing optimization-based methods, validating the generality and effectiveness of the proposed framework.

Problem

• spatial locality
• image editing
• policy optimization
• context distortion
• boundary inconsistency

Innovation

• locality preservation
• policy optimization
• image editing
• region-specific optimization
• spatial decoupling