Robust source coding scheme for image transmission over noisy-channels

The requirements of digital transmission systems are now becoming so severe that it is no longer possible to optimize independently the different functions of the system. In fact, a digital transmission system is composed of a source coding, a channel coding and modulation subsystems. The source coding sub-system aims at reducing the redundancy in the “source space” in order to improve the spectral efficiency of the transmission channel, and thus, to increase the number of transmitted sources in a given spectral band. The channel coding sub-system increases the resistance of the transmitted signal to channel errors. The modulation adapts the information to the transmission channel characteristics. In this study, the joint optimization of the source coding and the modulation sub-systems is discussed. The need of this optimization is shown to be of great importance in increasing error tolerance for channel impairments. The basic source coding scheme using the self organizing feature map (SOFM) neural network (performing vector quantization) is discussed. The paper presents the joint optimization of this source coding technique and the quadrature amplitude modulation (QAM). The simulated system and obtained results are discussed