Estimating the conditional entropy of a digital television signal

Conditional entropy (or conditional variable-length) coding has been proposed by various authors as a bit-rate reduction method for digital television transmission. Some claims for the effectiveness of the method were based on measurements of picture statistics using limited quantities of source data. The present paper points out the dangers of that approach and presents alternative measurements. Statistical experiments based on the new measurements are then described. The results provide an insight into the behaviour of the flawed conditional entropy estimates and indicate how they might be corrected, to assess more accurately the potential performance of conditional entropy coding.