Index Assignment for Two-Channel Quantization
János Balogh, János Csirik · IEEE Transactions on Information Theory · 2004
This paper concerns the design of a multiple description scalar quantization (MDSQ) system for two identical channels for an unbounded discrete information source. This translates to the combinatorial problem of finding an arrangement of the integers into the infinite plane square grid so that each row and each column contains exactly N numbers, such that the difference between any two numbers in the same row (or column) is at most d, with d to be minimized for a given N. The best previous lower and upper bounds on the lowest d were N/sup 2//3+O(N) and N/sup 2//2+O(N). We give new lower and upper bounds, both of the form 3N/sup 2//8+O(N). We also consider minimizing the maximal variance in any row or column and show that it must be at least N/sup 4//60+O(N/sup 3/), and that it does not have to be more than 3N/sup 4//160+O(N/sup 3/).