Are Watermarked Images Editable? SafeMark for Watermark-Preserving Text-Guided Image Editing

2026-05-19

AI Summary

This work addresses the challenge that text-guided image editing often compromises embedded watermarks, making it difficult to simultaneously preserve semantic editing quality and watermark integrity. The authors propose SafeMark, a novel framework that, for the first time, explicitly incorporates watermark recoverability into the optimization objective of diffusion-based editors. By introducing a thresholded differentiable watermark decoding loss, SafeMark protects watermark information without altering the model architecture. Grounded in mutual information analysis, the method provides an information-theoretic guarantee for robust watermark recovery after editing. Extensive experiments demonstrate that SafeMark consistently achieves high-quality semantic edits while maintaining high watermark bit accuracy across diverse datasets, editing techniques, and post-processing perturbations, effectively reconciling editing capability with watermark preservation.

Abstract

This paper investigates a fundamental yet underexplored question: can watermarked images remain editable without compromising watermark integrity? We propose SafeMark, a framework for watermark-preserving text-guided image manipulation that explicitly integrates watermark integrity into the editing process. Specifically, SafeMark adds a thresholded watermark-decoding loss directly to the diffusion editor's training objective, fine-tuning the editor so that semantically valid edits also preserve the embedded watermark at the final output. This design admits a clean information-theoretic justification: maintaining high bit-accuracy on the edited image lower-bounds the mutual information that the editor channel preserves between watermark and edited output, the quantity that fundamentally controls watermark recoverability. SafeMark is compatible with differentiable diffusion-based editors, and requires no architectural modification. Extensive evaluations across multiple datasets, text-guided editing methods, and post-edit distortion settings demonstrate that SafeMark achieves high watermark bit accuracy across diverse editing settings while maintaining high-quality semantic edits, without sacrificing robustness to common post-edit distortions. These results demonstrate that semantic editability and watermark integrity are fundamentally compatible, enabling trustworthy image provenance in generative editing pipelines.

Problem

• watermark integrity
• text-guided image editing
• image watermarking
• diffusion-based editing
• editable watermarked images

Innovation

• watermark-preserving editing
• diffusion-based image editing
• text-guided manipulation
• watermark integrity
• mutual information