MI-based Joint Source Channel Coding for Wireless Image Transmission
Ningwei Huang, Wenhui Hua, Lingyu Chen · 2023
Compared with traditional architecture of separated source-channel coding schemes, deep joint source-channel coding (D-JSCC) has advantages in having smaller delay and better robustness to fast channel variations. However, existing D-JSCC schemes lack the capability to optimize encoded channel input according to specific channel characteristics. To design for better adaptability for different communication channels, we propose a mutual-information-based joint source channel coding scheme, called M-JSCC. The scheme improves the encoding performance by training to maximize the mutual information between channel input and output. Experimental results show that the proposed M-JSCC scheme outperforms the conventional D-JSCC scheme in additive white Gaussian noise (AWGN) channels , Rayleigh fading channels and Rician channels for multiple image-quality metrics such as peak-signal-to-noise ratio (PSNR) and structural similarity (SSIM).