Performance Comparison of CNN and Transformer for JSCC in Semantic Image Transmission

This paper presents a comparative study of convolutional neural networks (CNNs) and Transformer-based architectures for joint source-channel coding(JSCC) in semantic image transmission. To assess the performance of both models, we adopt Gray-coded 16-QAM modulation over Rayleigh fading channels. Numerical results demonstrate that Transformer-based JSCC offers superior robustness in low signal-to-noise ratio (SNR) conditions, while CNN-based JSCC achieves better performance at high SNR levels.