Arbitrary source models and Bayesian codebooks in rate-distortion theory

We give a development of universal lossy data compression based on a lossy version of the Kraft inequality. As an application, we evaluate the performance of universal mixture codebooks.