Noisy channel performance of intraframe VQ coding of images

Results for vector quantization encoded data transmitted over a binary symmetric channel are presented. An empirical technique for predicting the effect of channel errors has been developed. To prevent unnecessary bandwidth expansion, a method of selectively protecting only the most vulnerable source bits was examined at two different bit-error probabilities. For low-rate source-coding application, the use of forward error-correcting coding resulted in a 3-5 dB signal-to-noise ratio improvement compared to systems using alternative source-coding rates but without error control.<>