Flex-DDCM adapts zero-shot image compression with diffusion guidance
Original title: Flex-DDCM: Zero-Shot Task-Oriented Image Compression via Diffusion Trajectory Guidance
Task guidance is embedded in the compressed trajectory without extra decoding complexity.
Flex-DDCM is a zero-shot task-oriented image compression framework built on Turbo-DDCM. It adapts the coded diffusion trajectory through task-guided noise selection so guidance effects sit in the compressed representation. Guidance terms can be combined, with timestep-dependent strengths during denoising. Experiments report trade-offs among pixel fidelity, perceptual quality, and downstream task performance, competitive with trained task-oriented codecs and faster decoding than Diff-ICMH.
- The method needs no compression-specific training.
- Guidance can be applied individually or combined.
- Decoding complexity does not increase from the task adaptation.
- Authors report faster decoding than Diff-ICMH.
From the original article
1 min read at arxiv.orgRecent zero-shot image compression methods exploit the powerful priors of pretrained diffusion models to produce perceptually pleasing reconstructions without compression-specific training. However, image compression serves not only human viewing but also machine analysis, while existing zero-shot codecs have limited flexibility in meeting the diverse reconstruction requirements of these applications.
By Yixin Gao · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.