移动通信2026，Vol.50Issue(6)：33-42,10.DOI:10.3969/j.issn.1006-1010.20260320-0001
面向通感融合的结构化生成式无线图像传输框架与关键技术
Structured Generative Wireless Image Transmission Framework for Integrated Sensing and Communication
摘要
Abstract
As wireless communications evolve towards intelligence and integrated sensing and communication(ISAC),realizing efficient and robust image transmission under limited bandwidth and unreliable channel conditions has become a critical challenge in intelligent wireless networks.However,existing wireless image transmission schemes primarily focus on enhancing visual quality and often lack active sensing and structured understanding of the source image content,such as the spatial consistency and saliency of targets,making it difficult to preserve the spatial layout of targets,salient regions,and high-level semantic consistency under low code-rate transmission and channel disturbances.To address these issues,this paper proposes a structured generative wireless image transmission framework oriented towards ISAC.At the transmitter,a sensing module performs saliency detection to extract a saliency map that characterizes the spatial layout of the scenario,serving as a sparse sensing prior.Subsequently,constrained by this structural sensing prior,a multimodal large language model(MLLM)is utilized to convert the image into structure-aware descriptive text,achieving a structured joint representation of semantics and sensing features.During the transmission phase,considering the importance and redundant characteristics of different modalities,modality-separated joint source-channel coding is employed to achieve low code-rate and anti-noise reliable transmission,where the saliency prior preserves key spatial structures and the text semantics convey high-level content information.At the receiver,the reconstruction process is modeled as a structured condition-guided generative workflow,utilizing the received sensing prior as an explicit constraint to guide high-quality image synthesis.Simulation results demonstrate that the proposed structured generative transmission framework significantly outperforms existing methods in terms of both semantic fidelity and visual perceptual quality,fully validating the superiority of ISAC and structured guidance in resource-constrained wireless image communications.关键词
无线图像通信/多模态大语言模型/联合信源信道编码/工作流/语义通信Key words
wireless image communication/multi-modal large language model/joint source and channel coding/semantic communication分类
信息技术与安全科学引用本文复制引用
王朔遥,叶凯浪,毕宿志,钱丽萍..面向通感融合的结构化生成式无线图像传输框架与关键技术[J].移动通信,2026,50(6):33-42,10.基金项目
国家自然科学基金项目"基于局部重传和自适码长的实时视频语义通信方法研究"(62571336) （62571336）
广东省基础与应用基础研究基金项目"面向二维与全景视频协同传输的边缘计算与码率自适应研究"(2025A1515010249) （2025A1515010249）
深大2035计划项目"面向实时视频通信的语义信息失真度量与信源网络信道联合编码研究"(2024C010) （2024C010）
广东省科学技术协会青年科技人才培育计划(SKXRC2025232) （SKXRC2025232）
