Leveraging Stable Diffusion with Context-Aware Prompts for Semantic Communication

Semantic communication in wireless image transmission leverages the meaning embedded in the image data, aiming to compress, transmit, and reconstruct images based on their semantic content rather than purely pixel data. This paradigm shift allows more efficient utilization of bandwidth and computational resources, focusing on extracting key features and contextual information that is critical for ensuring that the essential content of the image is preserved and accurately conveyed. In this study, we present a novel Stable Diffusion-based semantic communication (SDSC) framework that demonstrates high performance, characterized by an elevated bandwidth compression ratio (BCR) and robust noise tolerance achieved by diffusion mechanism integrating supplementary prompts. Our approach utilizes pre-trained modules of a Variational autoencoder (VAE) and a modified U-shaped network (UNet) to enable robust semantic encoding, decoding, and effective channel denoising. This scheme significantly enhances the system's ability to preserve data integrity and meaning in noisy environments. By introducing additional context-aware prompts during transmission, we improve the accuracy of received information and mitigate the adverse effects of interference and noise. Extensive simulations show that our framework outperforms previous innovative models, demonstrating superior communication fidelity and efficiency under various challenging conditions.