A database for data compression is universal if, in itsconstruction no prior knowledge of the source distribution is assumedand is optimal if, when we encode the reference index of the database,its encoding rate achieves the optimal encoding rate for any givensource: in the noiseless case the entropy rate and in the semifaithfulcase the rate-distortion function of the source. We construct auniversal database for all stationary ergodic sources, and prove theoptimality of the thus constructed database for a block-shift typereference and a single-shift type reference
展开▼