Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!hplabs!hpfcdc!bayes
From: bayes@hpfcdc.HP.COM (Scott Bayes)
Newsgroups: comp.graphics
Subject: Re: 8x8 dither matrices
Message-ID: <5040008@hpfcdc.HP.COM>
Date: 27 Apr 89 18:44:03 GMT
References: <6663@cbmvax.UUCP>
Organization: HP Ft. Collins, Co.
Lines: 30

A common way of enlarging a 4x4 matrix to 8x8 is:

multiply the matrix by the scalar 4.

tile the 8x8 matrix with 4 copies of the scaled-up 4x4 matrix

add the constant 1-matrix  	1 1 1 1
				1 1 1 1
				1 1 1 1
				1 1 1 1

to the upper right-hand tile. Add the constant 2-matrix to the lower left tile,
and add the constant 3-matrix to the lower right tile (actually you also add
the constant 0-matrix to the upper left tile). This produces an 8x8 matrix
that is fractally similar to the 4x4 starter, and shares many of its noise
distribution properties.

It can be applied to any 2^n by 2^n. For example n=0:

matrix = 0

Deriving the next up (n=1):

matrix =0 1
        2 3

etc.

Scott Bayes
dithering as usual