Deep JSCC

Deep Joint Source-Channel Coding (Deep JSCC) uses deep learning to optimize the simultaneous encoding and transmission of data, particularly images, over noisy channels, aiming for improved efficiency and perceptual quality compared to traditional separate source and channel coding methods. Current research focuses on incorporating generative models, such as diffusion models and variational autoencoders, to enhance image reconstruction fidelity and semantic preservation, especially under bandwidth or signal-to-noise ratio constraints. This approach holds significant promise for improving wireless image transmission in applications like 6G communication systems, offering potential gains in bandwidth efficiency and perceptual quality.