FreqEdit: Preserving High-Frequency Features for Robust Multi-Turn Image Editing

2025-12-01

AI Summary

In natural language instruction-driven multi-step image editing, repeated operations cause progressive degradation—particularly severe loss of high-frequency details—leading to distorted textures and identity drift. Method: This paper proposes a training-free stable editing framework. We first identify high-frequency attenuation as the primary cause of multi-step degradation and introduce three novel mechanisms: (i) diffusion-based high-frequency feature injection, (ii) reference velocity field-guided spatially adaptive modulation, and (iii) periodic editing-path compensation. Contribution/Results: The framework requires no additional training and effectively suppresses detail distortion and identity drift. Extensive experiments demonstrate that, over 10+ consecutive editing steps, our method outperforms seven state-of-the-art approaches in both identity preservation and instruction adherence, enabling high-quality, high-fidelity, controllable image evolution.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Instruction-based image editing through natural language has emerged as a powerful paradigm for intuitive visual manipulation. While recent models achieve impressive results on single edits, they suffer from severe quality degradation under multi-turn editing. Through systematic analysis, we identify progressive loss of high-frequency information as the primary cause of this quality degradation. We present FreqEdit, a training-free framework that enables stable editing across 10+ consecutive iterations. Our approach comprises three synergistic components: (1) high-frequency feature injection from reference velocity fields to preserve fine-grained details, (2) an adaptive injection strategy that spatially modulates injection strength for precise region-specific control, and (3) a path compensation mechanism that periodically recalibrates the editing trajectory to prevent over-constraint. Extensive experiments demonstrate that FreqEdit achieves superior performance in both identity preservation and instruction following compared to seven state-of-the-art baselines.

Problem

• Prevents quality loss in multi-turn image editing
• Preserves high-frequency details across editing iterations
• Enables stable and precise instruction-based visual manipulation

Innovation

• High-frequency feature injection from reference velocity fields
• Adaptive injection strategy for region-specific control
• Path compensation mechanism to prevent over-constraint