On the reduction of statistical dependencies in source signals
Abstract
Sampled source signals quantized to a finite set of possible levels (e.g., standard TV signals sampled at a rate of 10 MHz and quantized to 256 grey levels) are investigated. Such a source process, discrete in time and range, is redundant if the random variables are statistically dependent, or the distributions of the random variables are not uniform. Certain measures are defined, which permit to describe quantitatively the dependencies between the random variables of a process. The measures are discussed for transform coders and predictive coders. Some practical results are examined.
 Publication:

Frequenz
 Pub Date:
 April 1976
 DOI:
 10.1515/FREQ.1976.30.4.92
 Bibcode:
 1976Freq...30...92B
 Keywords:

 Data Reduction;
 Linear Transformations;
 Redundancy Encoding;
 Signal Encoding;
 Statistical Analysis;
 Stochastic Processes;
 Data Compression;
 Optimization;
 Pulse Code Modulation;
 Random Variables;
 Communications and Radar