Generative Semantic Communication via Alternating Dual-Domain Posterior Sampling

arXiv cs.CV / 4/21/2026

📰 NewsSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper studies generative semantic communication (SemCom) for wireless image transmission and argues that existing receiver designs based on MAP estimation can’t preserve the true data distribution, limiting perceptual quality.
It reformulates semantic decoding as a Bayesian inverse problem and shows that posterior sampling can achieve optimal perceptual quality by maintaining the data distribution.
The authors propose alternating dual-domain posterior sampling (ADDPS), a diffusion-based SemCom receiver that alternately enforces consistency in the latent domain and the image domain during sampling.
The alternating scheme is designed to decompose joint posterior sampling into easier subproblems, avoiding gradient conflicts while leveraging complementary benefits of both domains.
Experiments on FFHQ indicate that ADDPS delivers better perceptual quality than prior approaches.

Abstract

Generative semantic communication (SemCom) harnesses pretrained generative priors to improve the perceptual quality of wireless image transmission. Existing generative SemCom receivers, however, rely on maximum a posteriori (MAP) estimation, which fundamentally cannot preserve the data distribution and thus limits achievable perceptual quality. Moreover, current diffusion-based approaches using single-domain guidance face significant limitations: latent-domain guidance is sensitive to channel noise, while image-domain guidance inherits decoder bias. Simply combining both domains simultaneously yields an overconfident pseudo-posterior. In this paper, we formulate semantic decoding as a Bayesian inverse problem and prove that posterior sampling achieves optimal perceptual quality by preserving the data distribution. Building on this insight, we propose alternating dual-domain posterior sampling (ADDPS), a diffusion-based SemCom receiver that alternately enforces latent-domain and image-domain consistency during the sampling process. This alternating strategy decomposes joint posterior sampling into simpler subproblems, avoiding gradient conflicts while retaining the complementary strengths of both domains. Experiments on FFHQ demonstrate that the proposed ADDPS achieves superior perceptual quality compared with existing methods.