- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes EnDiC for jointly compressing a source image and its edited variants. It reuses diffusion trajectories for unchanged content and encodes local changes with adaptive bit allocation. Results show improved perceptual quality at low bitrates.
Critical strengths
- The problem is meaningful and practically relevant.
- Good perceptual results at low bitrates, with several recent baselines.
- Useful comparisons with region and residual coding.
Critical weaknesses
- The entity-oriented design mostly comes down to spatial masking. The novelty over existing reference-based coding is not very clear.
- Only 65 image clusters, all generated using one editing model and manually screened. Generalization remains unclear.
- The benefit of trajectory reuse is not sufficiently isolated from other components.
- No edited-region evaluation or computational cost analysis.
Questions for authors
- How much improvement comes specifically from trajectory reuse rather than masked direction coding?
- Does the method still work well with less spatially consistent edits?
- Can the authors report edited-region quality and encoding/decoding time?
2 comments
Authors’ response to Reviewer bWie (2/2)
W3 & Q1. The benefit of trajectory reuse is not sufficiently isolated from other components. How much improvement comes specifically from trajectory reuse rather than masked direction coding?
Thank you for raising this point. We add a controlled ablation in Section 4.4, Figure 8, comparing stepwise trajectory reuse with fusion only at the final latent or RGB output. All variants retain masked direction coding and share the same source and mask streams and per-step atom counts. We re-encode the edit directions for each control under this fixed allocation, yielding identical total bpp across variants. The representative low-rate results below show the additional gains from stepwise trajectory reuse under these controlled conditions.
Kodak
| Fusion | bpp | PSNR ↑ | LPIPS-V ↓ | DISTS ↓ |
|---|---|---|---|---|
| Final RGB | 0.0170 | 22.602 | 0.2967 | 0.1143 |
| Final latent | 0.0170 | 22.718 | 0.2911 | 0.1135 |
| Stepwise (EnDiC) | 0.0170 | 22.817 | 0.2846 | 0.1074 |
CLIC512
| Fusion | bpp | PSNR ↑ | LPIPS-V ↓ | DISTS ↓ |
|---|---|---|---|---|
| Final RGB | 0.0179 | 22.457 | 0.2723 | 0.1163 |
| Final latent | 0.0179 | 22.632 | 0.2666 | 0.1136 |
| Stepwise (EnDiC) | 0.0179 | 22.715 | 0.2626 | 0.1096 |
W4 & Q3. No edited-region evaluation or computational cost analysis. Can the authors report edited-region quality and encoding/decoding time?
Thank you for pointing this out. We have added both regional reconstruction evaluation and computational cost analysis to the paper.
Edited-region quality. We add separate PSNR and LPIPS evaluations for edited and unchanged regions in Section 4.3 (Figure 7), with metric definitions in Appendix A.1. We also add source-image reconstruction results in Appendix A.4.
The table below provides a preview. At matched total rates, EnDiC achieves higher PSNR and lower LPIPS than independent Turbo-DDCM for the source, edited regions, and unchanged regions on both datasets. Total bitrate includes the source, all three edits, masks, and required side information.
ImaGeneSet-Kodak, total bpp = 0.03377
| Region | Method | PSNR ↑ | LPIPS ↓ |
|---|---|---|---|
| Source | Turbo-DDCM | 22.65 | 0.3099 |
| Source | EnDiC | 25.14 | 0.2019 |
| Edited region | Turbo-DDCM | 21.79 | 0.2827 |
| Edited region | EnDiC | 24.67 | 0.1871 |
| Unchanged region | Turbo-DDCM | 21.73 | 0.3260 |
| Unchanged region | EnDiC | 23.76 | 0.2332 |
ImaGeneSet-CLIC512, total bpp = 0.05860
| Region | Method | PSNR ↑ | LPIPS ↓ |
|---|---|---|---|
| Source | Turbo-DDCM | 23.55 | 0.2391 |
| Source | EnDiC | 26.08 | 0.1512 |
| Edited region | Turbo-DDCM | 22.17 | 0.2361 |
| Edited region | EnDiC | 24.66 | 0.1667 |
| Unchanged region | Turbo-DDCM | 22.65 | 0.2540 |
| Unchanged region | EnDiC | 24.67 | 0.1835 |
Encoding and decoding time. We have added computational cost analysis and the timing protocol in Appendix A.10. All measurements are performed on an NVIDIA RTX 4090, with mask generation excluded. The following table summarizes the encoding and decoding times per four-image cluster. EnDiC's encoding time includes the optional 80-second refinement.
ImaGeneSet-Kodak
| Method | Encoding (s) | Decoding (s) |
|---|---|---|
| EnDiC () | 107.64 | 6.43 |
| EnDiC-Fast () | 8.08 | 6.45 |
| Turbo-DDCM (independent, ) | 6.91 | 6.63 |
ImaGeneSet-CLIC512
| Method | Encoding (s) | Decoding (s) |
|---|---|---|
| EnDiC () | 120.35 | 6.69 |
| EnDiC-Fast () | 8.29 | 6.65 |
| Turbo-DDCM (independent, ) | 7.06 | 6.78 |
EnDiC-Fast reduces encoding time by on Kodak and on CLIC512, with minor quality degradation and nearly unchanged decoding time. For reference, Turbo-DDCM independently encodes the four images at .
Authors’ response to Reviewer bWie (1/2)
Thank you for your careful review and constructive feedback. We appreciate your time and consideration and address your comments and questions below.
W1. The entity-oriented design mostly comes down to spatial masking. The novelty over existing reference-based coding is not very clear.
1. Joint coding beyond spatial masking. We study ImaGene Compression, a new setting where a source image and its edited variants share spatially aligned content and differ in localized entity changes. Existing reference-based codecs typically establish correspondence with a reference and inject its information into the target representation. EnDiC instead restructures the transmitted information and decoding process: the source trajectory is encoded once, while the edit masks and incremental direction codes guide the diffusion decoder to reuse source states and reconstruct entity changes throughout denoising. RSBA further adapts edit budgets to change-region extent and allocates bits across denoising steps according to reconstruction benefit.
2. Added comparison with reference-based coding. We add a comparison with CLC (AAAI 2025) in Appendix A.7, Figure 18. CLC compresses individual images using retrieved references. We additionally supply the decoded source as a reference candidate for each edit in CLC. EnDiC achieves competitive LPIPS and DISTS at lower bitrates on both datasets.
W2a. Only 65 image clusters, all generated using one editing model and manually screened. Generalization remains unclear.
Our dataset size and source-image selection follow common practice in image compression evaluation. We use all images from two standard compression benchmarks, Kodak (24 images) and the CLIC2020 Professional validation set (41 images), covering natural scenes, buildings, people, and everyday objects. Each source has three edited variants generated using Seedream 5.0 Pro, covering entity addition, removal, and modification. We add detailed clarification of the existing screening procedure in Appendix A.1, where edits are selected for instruction adherence, visual plausibility, and preservation of non-target content. In future work, we plan to expand ImaGeneSet with more source images and a broader range of image editing methods.
W2b & Q2. Does the method still work well with less spatially consistent edits?
We add visualizations in Appendix A.2 (Figure 12) showing subtle color, texture, and alignment differences that remain outside the detected masks. Thus, our evaluated source-edit pairs are not perfectly spatially consistent.
For additional changes detected by semantic localization or RGB-difference cues, EnDiC includes the affected regions in the edit mask and encodes them. Larger coded regions leave less shared content to reuse, reducing bitrate savings.
Large viewpoint changes and geometric transformations remain outside our current evaluation. Evaluating and extending EnDiC to such edits is an interesting direction for future work.