Super-Resolution Semantic Communication System for Satellite Image

This paper introduces SR-DeepSC, a novel super- resolution semantic communication system to serve for satellite image transmission. We propose an asymmetric architecture utilizing Vision Mamba-inspired blocks, tailored for both satellite and ground station environments. SR-DeepSC significantly outperforms traditional methods like JPEG+Cubic in image quality, maintaining high SSIM and PSNR values across various SNR levels. The system's lightweight satellite-side implementation and more complex ground station processing effectively balance computational load, reducing latency while improving reconstruction quality. Experimental results demonstrate SR-DeepSC's robustness to noise and efficiency in bandwidth-limited conditions, making it a promising solution for enhanced satellite image communication.