Path: utzoo!attcan!uunet!timbuk!cs.umn.edu!msi-s0.msi.umn.edu!srcsip!sol.ctr.columbia.edu!samsung!usc!apple!bionet!agate!eos!jbm
From: jbm@eos.UUCP (Jeffrey Mulligan)
Newsgroups: comp.graphics
Subject: Re: dithering
Keywords: grayscale to black&white
Message-ID: <7312@eos.UUCP>
Date: 27 Sep 90 18:43:46 GMT
References: <13919@netcom.UUCP>
Organization: NASA Ames Research Center, California
Lines: 27

allanh@netcom.UUCP (Allan N. Hessenflow) writes:

>I'm looking for any references on dithering algorithms that work for an output
>device with higher addressable resolution than dot-size.  Specifically, the
>dots are round and larger than 1/300 inch in diameter.  The addressable
>resolution is 300dpi vertically, and several times that horizontally.

>I can pretend the dots are square with 1/300 inch sides and only use 1/300 inch
>horizontal resolution with reasonable results, although they aren't very linear
>because adjacent dots overlap.  I'm hoping to find a method which can take
>advantage of the high horizontal resolution and deal with the fact that the
>dots overlap.

A "clustered dot" algorithm is generally indicated when you have overlapping
printer dots, because it at least does a pretty good job of insuring
monotonicity.  As you point out, the main problem is linearity.
My suggestion would be to dither a gray scale ramp, measure the
resulting non-linearity, invert this function and use it to
transform your input imagery prior to dithering.  The fact that your
device has anisotropic resolution means that you will want to
design a non-square dithering matrix.

-- 

	Jeff Mulligan (jbm@eos.arc.nasa.gov)
	NASA/Ames Research Ctr., Mail Stop 262-2, Moffet Field CA, 94035
	(415) 604-3745