Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!zephyr.ens.tek.com!gvgpsa!gold.gvg.tek.com!johna
From: johna@gold.gvg.tek.com (John Abt)
Newsgroups: comp.multimedia
Subject: Re: Quantization & dithering definitions please?
Message-ID: <2399@gold.gvg.tek.com>
Date: 30 May 91 15:39:11 GMT
References: <1991May30.093848.8487@cc.curtin.edu.au>
Organization: Grass Valley Group, Grass Valley, CA
Lines: 28

In article <1991May30.093848.8487@cc.curtin.edu.au> chill@cc.curtin.edu.au writes:
>I've been listening to tapes from the 91 Apple Worldwide Developers Conference
>in the area of audio-video integration, and I'm floundering soemwhat with
>some of the new jargon terms.
>Can someone explain in words of few syllables what:
>quantization and dithering are please?

Quantization is a process which produces finite numerical values at
a given resolution from inputs of arbitrary (but greater) resolution. 
Rounding-off or truncation are examples of quantizing, as is analog-
to-digital conversion. 
   A problem with quantization is correlated error. When a correlated 
input is quantized, the error resulting from the quantization operation 
is also correlated. Correlated error tends to "stand-out" and is therefore 
objectionable. Contour lines in an image gradient are a good example of 
correlated error. "Dithering" is a process which attempts to "un-correlate" 
quantization error. Essentially, a "dither" is a noise signal which is 
added prior to quantizing. This has the effect of "breaking-up" the 
patterns of correlated error. In the image example above, a contour line 
exists because pixels "make-the-jump" along a distinct spatial line. 
Dithering causes the line to be less distinct by moving it around a little.
   Although dithering is the subject of considerable research and 
development, the concept certainly is not new. A good treatment can be 
found in "Television by Pulse Code Modulation", W.M. Goodall, Bell System 
Technical Journal, pp. 33-49, January 1951.

John Abt
Grass Valley Group