PERFORMANCE OF ENTROPY-CONSTRAINED BLOCK TRANSFORM QUANTIZERS.

There are numerous techniques available for analog-to-digital conversion, or data compression, of discrete-time processes. They range from simple scalar (zero-memory) quantization or PCM, to more sophisticated schemes such as predictive encoding, tree encoding and multi-dimensional quantization. One of the most popular data compression schemes for encoding of correlated sources is block transform quantization. The main advantages associated with the block transform quantization scheme are: (i) good performance, and (ii) ease of implementation. We study the rate-distortion theoretic performance of optimal block transform quantization schemes on first-order stationary autoregressive processes. More precisely, we assume that blocks of length L of source outputs are operated upon by the Karhunen-Loeve transformation. Only summary is presented.