Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!calvin.usc.edu!alves
From: alves@calvin.usc.edu (William Alves)
Newsgroups: comp.multimedia
Subject: Re: Quantization & dithering definitions please?
Message-ID: <33318@usc>
Date: 4 Jun 91 01:53:17 GMT
References: <1991May30.093848.8487@cc.curtin.edu.au>
Sender: news@usc
Organization: University of Southern California, Los Angeles, CA
Lines: 69
Nntp-Posting-Host: calvin.usc.edu

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?
>
As another poster said, quantization is simply assigning discrete numbers
to analog values. It's usually used in reference to analog-to-digital
conversion.

There are a number of problems (or "artifacts" in the jargon) which occur
as a byproduct of quantization and sampling (assigning these numbers at
regular points in space or time). In spatial sampling a common artifact
is jagged edges, in color visible contours or "bands," and in time 
jerkiness of motion. These artifacts are all analogous in the ways that
our senses recognize repetitive patterns and are even drawn to them. In
signal processing language, these are results of localized high frequen-
cies. Our eyes will not notice them as much if they are filtered out and
their energy distributed over a wider spectrum.

Dither is a relatively easy "filter" of this type to implement. A small
amount of "noise" or randomization is injected into the signal before 
quantization. For example, if color of pixels in the area around a sudden
change of color is slightly randomized, the eye will tend to "average"
the colors and see a smoother transition (and less jagged edges and visi-
ble contours).

The first important investigations of dither were in television. Some
references include:
L. G. Roberts, "Picture Coding Using Pseudorandom Noise," IRE Transactions in
Information Theory IT-8, Feb. 1962, 145-154.
 L. Schuchman, "Dither Signals and Their Effect on Quantization," IEEE
Transactions on Commcation Theory, COM-12 (Dec. 1962), 162-165.
 J.F. Jarvis, N. Judice, and W.H. Ninke, "A Survey of Techniques for the
Display of Continuous Tone Pictures on Bilevel Displays," Computer Graphics
and Image Processing 5, No. 1 (1976), 13-40.

Dither is also used in digital audio. Analogously to the eye's attraction
to repetitive structures, the ear tends to find harmonic distortion much
more objectionable than the equivalent amount of noise. But at very low
amplitudes, where the signal is represented only by a bit or two, the sig-
nal can be greatly distorted. However, if a small amount of noise is in-
jected into the signal before the analog-to-digital conversion, the energy
of the objectionable harmonics are in effect spread over the entire spectrum.

For references here, see:

 Ken C. Pohlmann, Principles of Digital Audio, (Indianapolis, IN: Howard W.
Sams & Co., 1985), 58-59.
 J. Vanderkooy and S.P. Lipschitz, "Resolution Below the Least Significant Bit
in Digital Audio Systems with Dither," Journal of the Audio Engineering
Society, March 1984.
 Barry A. Blesser, "Digitization of Audio: A Comprehensive Examination of
Theory, Implementation, and Current Practice," Journal of the Audio Engineering
Society, 26, No. 10 (Oct. 1978) 739-771.

Often, one also sees dither used to refer to similar non-random processes.
For example, "dither" on some scanner software I have used actually puts a
patterned alternation between two colors to simulate a color in between.
Because of the regular pattern, this is not really dither, at least as I
have defined it. While it will help simulate extra colors, the repetition
of the pattern introduces a new artifact which catches the eye.

I hope I didn't use too many long words, and I hope this helps.

Bill Alves