Joint Source/Channel Coding of Multispectral Imagery

"In this paper we investigate the problem of compressing and reliably transmitting multispectral imagery over binary symmetric channels. A novel source/channel multispectral image coding scheme is proposed. The basic engines of our scheme is a powerful de-correlating transform both in spectral and spatial domains, an optimal bit allocation procedure, block classification and a robust quantizer to cope with the noise introduced by the channel. SNR results show that our coder is very competitive with other approaches when the channel bit error rate is zero, while still attaining robustness to channel errors when the channel is noisy. Furthermore, graceful degradation is also observed as the channel cross increases."