Simulated annealing for robust VQ: improving image transmission through a fading channel

Abstract

Vector quantization (VQ) has been extensively used in image coding systems. However, when the communications system involves a noise channel, it is well known that VQ is highly sensitive to channel errors. As a consequence, a considerable degradation may be introduced in the reconstructed images. In the present paper, simulated annealing (SA) is applied to robust VQ, that is, it is used for assigning binary indexes to the VQ code vectors, as an attempt to circumvent the effect of channel errors. Simulations involving image transmission through a Rayleigh fading channel clearly indicate the suitability of SA. The authors also show that an additional improvement can be obtained by using an alternative initialization of the SA algorithm.