From albanycs!leah:rsb584 Mon Jan 18 16:56:40 1988
Received: by albanycs.albany.edu (5.54/4.8)
	id AA03293; Mon, 18 Jan 88 16:04:51 EST
Date: Mon, 18 Jan 88 16:04:48 EST
From: albanycs!leah:rsb584 ( Raymond S Brand)
Received: by leah.Albany.EDU (5.58/1.1)
	id AA17806; Mon, 18 Jan 88 16:04:48 EST
Message-Id: <8801182104.AA17806@leah.Albany.EDU>
To: albanycs:beowulf!rsbx
Subject: dither6.txt

>From lou@aramis.rutgers.edu Mon Jan 11 00:02:13 1988
Path: leah!uwmcsd1!ig!jade!ucbvax!husc6!rutgers!aramis.rutgers.edu!lou
From: lou@aramis.rutgers.edu (Lou Steinberg)
Newsgroups: comp.graphics
Subject: Re: Advanced Dither Needed
Keywords: dither
Message-ID: <2709@aramis.rutgers.edu>
Date: 11 Jan 88 05:02:13 GMT
References: <3703@ames.arpa> <2741@masscomp.UUCP>
Distribution: na
Organization: Rutgers Univ., New Brunswick, N.J.
Lines: 29

In article <2741@masscomp.UUCP> garyo@masscomp.UUCP (Gary Oberbrunner) writes:
> But the wayd [Floyd/Steinberg] dither works is like this (it's so simple...)
  [...]
> Add 3/8 of that error to the pixel to the right, and 3/8 to the pixel
> below the current pixel.  The remaining 1/4 goes to the pixel diagonally
> below.

I should point out that the actual fractions we used were, assuming
you are at X, moving left to right:

		X     7/16
         3/16   5/16  1/16    

Note that the error goes to four neighbors, not three.  I think this
will probably do better (at least for black and white) than the
3/8-3/8-1/4 distribution, at the cost of greater processing.  I have
seen the 3/8-3/8-1/4 distribution described as "our" algorithm before,
but I have no idea who the credit really belongs to.

Also, I should add that if you do zig-zag scanning (see my immediately
previous message), it is sufficient (but not quite as good) to send
half the error one pixel ahead (e.g. to the right on lines you scan
left to right), and half one pixel straight down.  Again, this is for
black and white;  I've not tried it with color.
-- 
					Lou Steinberg

uucp:   {pretty much any major site}!rutgers!aramis.rutgers.edu!lou 
arpa:   lou@aramis.rutgers.edu