This paper presents an analysis of the effects of statistically independent channel errors on the mean-squared error performance of image transform coding systems. The analysis is discussed for several different stochasic statistics of the quantizer input valuse. The stochastic distributions under consideration here are Laplacian, Gaussian and uniform. For each case of the distributions, we evaluate the MSE performance when NBC, FBC, MDC and Gray code respectively is employed for encoding the quantized values of the transformed coeffecients into the corresponding code words. The result of this study shows that what code is desired to be chosen to minimize the MSE for the given stochastic distributions of the quantizer input values.