SING: semantic image communications using null-space and INN-guided diffusion models
File(s)
Accepted version
Author(s)
Chen, Jiakang
Yilmaz, Selim F
You, Di
Dragotti, Pier Luigi
Gunduz, Deniz
Type
Journal Article
Abstract
Joint source-channel coding systems based on deep neural networks (DeepJSCC) have recently achieved remarkable performance in wireless image transmission. Existing methods primarily minimize pixel-level distortion between the transmitted image and its reconstruction at the receiver, often overlooking perceptual quality. This can lead to severe perceptual degradation when transmitting images under extreme conditions, such as low bandwidth compression ratios (BCRs) and low signal-to-noise ratios (SNRs). In this work, we propose SING, a novel two-stage JSCC framework that formulates the recovery of high-quality source images from corrupted reconstructions as an inverse problem. Depending on the prior knowledge of the DeepJSCC encoder/decoder and channel, SING either approximates the stochastic degradation by a linear proxy (SING-Zero), or learns a nonlinear proxy using invertible neural networks (INNs) (SING-INN). Both approaches enable the seamless integration of pretrained unconditional diffusion models as a generative prior, enhancing perceptual quality. Experimental results demonstrate that SING outperforms DeepJSCC and other approaches, delivering superior perceptual quality even under extremely challenging conditions, including scenarios with significant distribution mismatches between the training and test data.
Date Issued
2026-09-29
Date Acceptance
2026-09-17
Citation
IEEE Transactions on Communications, 2026, 74, pp.15092-15108
ISSN
0090-6778
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Start Page
15092
End Page
15108
Journal / Book Title
IEEE Transactions on Communications
Volume
74
Copyright Statement
Copyright © 2026, IEEE. This is the author’s accepted manuscript made available under a CC-BY licence in accordance with Imperial’s Research Publications Open Access policy (www.imperial.ac.uk/oa-policy)
License URL
Publication Status
Published
Date Publish Online
2026-09-29