On a universal antidictionary coding for stationary ergodic sources with finite alphabet

This paper shows that a two-pass universal antidictionary coding method is asymptotically optimal for stationary ergodic sources with a finite alphabet. To prove the results, we propose a new compact tree representation of an antidictionary. We also extend the lossless compression algorithm proposed by Dubé and Beaudoin, called CSE, from the binary alphabet to the q-ary alphabet (q ≥ 2), which we utilize to efficiently compress the antidictionary of an input sequence to be encoded.