Diffusion-OAMP for Joint Image Compression and Wireless Transmission

Joint image compression and wireless transmission is a critical but underexplored area in practical communication systems. To address this, researchers have introduced Diffusion-OAMP, a novel framework that formulates the problem under an equivalent linear model. The key innovation is integrating a pre-trained diffusion model into the Orthogonal Approximate Message Passing (OAMP) algorithm without requiring additional training. In this setup, the OAMP linear estimator produces pseudo-AWGN (Additive White Gaussian Noise) observations, and the diffusion model acts as a nonlinear estimator guided by an SNR-matching rule. This hybrid approach allows multiple generative priors to be incorporated seamlessly into OAMP, potentially improving robustness in real-world channels.

The method was tested across various compression ratios and noise levels, outperforming classic methods in all evaluated settings. The results, detailed in a 6-page paper with 5 figures and 2 tables, demonstrate that Diffusion-OAMP can effectively balance compression efficiency and transmission reliability. Notably, because the framework is training-free and relies on a pre-trained diffusion model, it can adapt to different channel conditions without costly retraining. This opens up new possibilities for deploying generative models in communication systems, especially where bandwidth and noise are unpredictable. The paper is available on arXiv and has been submitted for publication.