Rate and distortion redundancies for universal source coding with respect to a fidelity criterion

Rissanen has shown that there exist universal noiseless codes for {Xi} with per-letter rate redundancy as low as (K log N)/2N, where N is the blocklength and K is the number of source parameters. we derive an analogous result for universal source coding with respect to the squared error fidelity criterion: there exist codes with per-letter rate redundancy as low as (K log N)/2N and per-letter distortion (averaged over X^N and θ) at most D(R)[1 + K/N], where D(r) is an average distortion-rate function and K is now the number of parameters in the code.

[1]  Jorma Rissanen,et al.  Universal coding, information, prediction, and estimation , 1984, IEEE Trans. Inf. Theory.

[2]  Sethuraman Panchanathan,et al.  Algorithms And Architecture For Image Adaptive Vector Quantization , 1988, Other Conferences.

[3]  Kenneth Zeger,et al.  Universal adaptive vector quantization using codebook quantization with application to image compression , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[4]  Anurag Bist,et al.  of Book , 2022 .