Semantic Communication based on Large Language Model for Underwater Image Transmission

TL;DR

This work tackles the bottleneck of underwater image transmission by introducing an LLM-based semantic communication framework that leverages visual LLMs to identify and prioritize semantically important regions for transmission. The transmitter performs semantic encoding with LLM-driven prioritization and a highly compressed full image, while the receiver uses a diffusion-based decoder guided by two specialized ControlNets and an LLM-based textual recovery to reconstruct high-fidelity images under noisy acoustic channels. Key contributions include the integration of LLM-based information prioritization, two ControlNet networks, and a text-guided diffusion recovery, achieving substantial data-size reductions while preserving semantic content, demonstrated on the SUIM dataset with multiple metrics (FID, SSIM, CLIP, LPIPS) across varying SNRs. The results indicate improved semantic alignment and perceptual quality over baselines, highlighting the practical potential for robust, efficient underwater sensing and monitoring in bandwidth-constrained environments.

Abstract

Underwater communication is essential for environmental monitoring, marine biology research, and underwater exploration. Traditional underwater communication faces limitations like low bandwidth, high latency, and susceptibility to noise, while semantic communication (SC) offers a promising solution by focusing on the exchange of semantics rather than symbols or bits. However, SC encounters challenges in underwater environments, including semantic information mismatch and difficulties in accurately identifying and transmitting critical information that aligns with the diverse requirements of underwater applications. To address these challenges, we propose a novel Semantic Communication (SC) framework based on Large Language Models (LLMs). Our framework leverages visual LLMs to perform semantic compression and prioritization of underwater image data according to the query from users. By identifying and encoding key semantic elements within the images, the system selectively transmits high-priority information while applying higher compression rates to less critical regions. On the receiver side, an LLM-based recovery mechanism, along with Global Vision ControlNet and Key Region ControlNet networks, aids in reconstructing the images, thereby enhancing communication efficiency and robustness. Our framework reduces the overall data size to 0.8\% of the original. Experimental results demonstrate that our method significantly outperforms existing approaches, ensuring high-quality, semantically accurate image reconstruction.

Figures (6)

• Figure 1: The framework structure of proposed LLM-based semantic communication.
• Figure 2: The testing evaluation of different approaches.
• Figure 3: The generation picture of different approaches when SNR=0.
• Figure 4: The testing evaluation of ablation study.
• Figure 5: The testing evaluation of different $\alpha$.