New Results on Variable-Length Lossy Compression Allowing Positive Overflow and Excess Distortion Probabilities

Abstract

This paper shows some new results for the problem of variable-length lossy source coding. We deal with the case where both the excess distortion probability and the overflow probability of codeword lengths are less than or equal to positive constants. Our previous study for the problem of variable-length (noiseless) lossy source coding has derived the general formula of the infimum of the thresholds on the overflow probability by using the quantity based on the smooth max entropy. This study extends this result in two directions. First, we derive the single-letter characterization of the infimum of the thresholds on the overflow probability for stationary memoryless sources. Second, for the problem of variable-length noisy lossy source coding, also known as the problem of remote lossy source coding, we establish the general nonasymptotic formula on the converse bound by using the new quantity based on the smooth max entropy.