Title :
Optimum quantization of order statistics
Author :
Gandhi, Prashant P.
Author_Institution :
Adv. Audio Design, Texas Instrum. Inc., Dallas, TX, USA
fDate :
9/1/1997 12:00:00 AM
Abstract :
We present necessary conditions for minimum mean square error (MMSE) quantization of order statistics. These conditions are obtained by observing that in a quantize-sort cascade system, operations of quantization and sorting are interchangeable. We consider MMSE quantization of a single order statistic as well as of a group of order statistics and highlight several interesting properties. Numerical characteristics of order statistic quantizers are presented for Gaussian and uniform distributions, and detailed comparisons are made
Keywords :
Gaussian distribution; error statistics; optimisation; quantisation (signal); signal sampling; statistical analysis; Gaussian distribution; MMSE quantization; analog samples; digital signal processing; group order statistics; minimum mean square error; necessary conditions; numerical characteristics; optimum quantization; order statistic quantizers; quantize-sort cascade system; single order statistics; sorting; uniform distribution; Error analysis; Filters; Focusing; Image processing; Mean square error methods; Quantization; Signal processing; Sorting; Statistical distributions; Statistics;
Journal_Title :
Signal Processing, IEEE Transactions on
