Adaptive Semantic Image Transmission Using Generative Foundation Model

Semantic communication has shown promise in improving data transmission efficiency under critical scenarios, particularly through the use of foundation models (FMs). This study introduces FMSC, a semantic communication framework based on FMs. The framework employs FM-based segmentation and reconstruction to significantly reduce bandwidth requirements and accurately recover semantic features in noisy and interfered environments. To address varying scenarios, an adaptive encoder-decoder is proposed to protect important features and satisfy user requirements. Particularly, the proposed method can leverage a well-received image as a reference for repairing damaged images.