08/01/1990 09:00 AM

Computer Science

Results of Gallager and VanVoorhis are employed to show that all augend-based binary arithmetic coders in the fixed augend case generate code strings of identical length to some Golomb code and vice-versa. A-based coders are studied for the random augend case (random interval model) that assume a randomization of successive augend values. When augend Q and the smaller binary probability q_{bpd} of the stationary binary input sequence are *matched*, then the best coding efficiency is achieved. The matching values of Q/q_{opt} differ for fixed and random augend cases, but converge as q becomes small. The coding efficiencies are also compared. The Golomb code matching point for best efficiency is the basis for Golomb number m.