Multiple-description image transmission over wireless channels

This thesis presents several image transmission methods over wireless channels utilizing multiple description coding (MDC). The MDC principle is used as a means of providing a solution for the problems of multimedia transmission over unreliable and potentially catastrophic channels. Cases of particular interest involve wireless channels, over which data packets are received with the long error bursts due to fading, shadowing, or occasional link failures.; Multiple description coding was originally developed for operation in wireline networks in which a system may be modeled to have multiple independent on-off channels between the transmitter and the receiver. We replace the parallel independent on-off channel model with the Gilbert-Elliott channel model, which considers the bursty characteristics of erasures during deep fading/shadowing in a wireless link. In this case, we have proposed a two-description MDC method to operate in a two-state Markov channel model.; We present a combined method of MDC and space-time block coding (STBC) for multiple transmit antenna systems. In this work, we employ Alamouti's transmit diversity scheme, with a new transmit sequence, without any loss of bit rate, which can break the superimposed wireless channel. Thus, the MDC could be applied to the bit error wireless channel model as well as an erasure channel model. The proposed method provides source diversity as well as channel diversity due to the combined use of MDC and space-time block coding (STBC).; The unequal error protection (UEP) property of multi-resolution (MR) modulation is then combined with MDC for image transmission. The multi-level constellation in the MR modulation introduces an additional channel state that allows unequal importance to be placed on the descriptions. This provides the ability to recover the transmitted image from partial information. The constellation points in the MR modulation are made to adapt to the channel state and mobile station information.