Image/video communications: joint source/channel coding

Transmission of image/video messages over communication networks is becoming a standard way of communication due to very efficient compression algorithms that reduce required channel capacity to an acceptable level. However, all compression standard techniques are strongly sensitivitive to channel disturbances and their application is suitable only for practically noiseless channels. In standard noisy channels, the effect of errors on a compressed data bit stream can be divided into two categories: systematic errors defined by the structure of data blocks, and random errors caused by amplitude changes of transmitted components. A systematic error can be detected at the receiver through control of the data stream structure and corrected by error concealment methods or by automatic repeat request (ARQ) procedures. Random errors, noise and burst-like errors, as well as impulse noise, should be controlled through channel coding. It is reasonable that an integrated source and channel coding methods should be preferred and should give better coding performance. In this paper a new framework for an image/video coding approach has been presented in which the source and channel coding is integrated in a unique procedure. Image compression is performed in a standard way of the JPEG algorithm based on discrete cosine transform (DCT) and error control coding is based on the real/complex-number (N, M) BCH code using discrete Fourier transform (DFT) specified with zeros in the time domain i.e. with roots in the frequency domain. Efficiency of the proposed method is tested on two examples, an one-dimensional real-valued time sequence coded by real-number (20, 16) BCH code using DFT, and an example of an image coded by complex (10, 8)BCH code using DFT with the correction ability of up to 8 impulses per transmitted 8x8 block. In addition, two decoding methods based on Berlekamp-Massey algorithm (BMA) and the minimum-norm algorithm (MNA) have also compared.