Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ames!eos!jbm
From: jbm@eos.UUCP (Jeffrey Mulligan)
Newsgroups: comp.graphics
Subject: Re: Extended dithering, any work done?
Message-ID: <3362@eos.UUCP>
Date: 25 Apr 89 20:53:26 GMT
References: <5237@hubcap.clemson.edu>
Organization: NASA Ames Research Center, California
Lines: 31

From article <5237@hubcap.clemson.edu>, by jcocon%hazel.cs.clemson.edu@hubcap.clemson.edu (James C O'Connor III, 2846):

> Has any work been done on displaying 256 grey shades on a monitor with
> greater than two grey shades (i.e. black and white + some number of greys)?
> It seems that an extension to dither would be a good place to start.

Some time ago I wrote a generalization of the F-S error diffuser to
use multiple quantization levels (the values of the levels could
be user-defined).  Then just a couple of weeks ago, when trying to
get a multiple level ordered dither algorithm that would run on an
array processor, I discovered that there is a very simple way to do
this which only involves 1 bit dithering:

Say the input image is already quantized, from 0 to N-1 (N levels),
and that you want to re-quantize to go between 0 and M-1 (M levels).
Let's say that the final M levels should
be uniformly spaced on the same range as the input,
so the increment between levels will be (N-1)/(M-1).
Begin by just truncating the input to M-1 levels by doing an integer
division by M-1, and multiplying the result by (N-1)/(M-1).  Then compute
the error image by subtraction.  Now dither the error image down
to 1 bit, and combine with the original quantized image.

I have left out a little detail about scaling to avoid getting M+1 bits in
the result, but that is the gist of it.

-- 

	Jeff Mulligan (jbm@aurora.arc.nasa.gov)
	NASA/Ames Research Ctr., Mail Stop 239-3, Moffet Field CA, 94035
	(415) 694-6290