Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.csd.uwm.edu!cs.utexas.edu!uunet!mcsun!unido!isaak!leonardo!schwarze
From: schwarze%leonardo@isaak.uucp (Jochen Schwarze)
Newsgroups: comp.graphics
Subject: Re: How to map 24-bit RGB to 256 co
Message-ID: <1412@isaak.UUCP>
Date: 3 Sep 89 11:38:37 GMT
References: <2009@uceng.UC.EDU> <5300026@ux1.cso.uiuc.edu>
Sender: news@isaak.UUCP
Reply-To: schwarze%leonardo@isaak.UUCP (Jochen Schwarze)
Organization: ISA GmbH, Stuttgart, West-Germany
Lines: 31
Bcc: schwarze


Well, this might have been dicussed before and I don't know if it
helps for the topic, but anyway:

To map a large number of gray scale levels to a considerable smaller
one (say 65536 to 64 or so) I used an algorithm that numerically
integrates the gray scale histogram of the image and uses the
integrated table to map from the large number of levels to the small
one. In pseudo code:

    int histogram[ 0..65535 ]

    foreach pixel
	++histogram[ pixel_value ]

    foreach i from 1..65535
	histogram[ i ] += histogram[ i - 1 ]

The histogram table now contains a growing series (don't know the
exact term in English) of numbers. It has its biggest slope, i.e the
finest resolution in that range of pixel values that mainly occur in
the image. The histogram table can then be scaled down by a factor
like 64 / num_pixels (if 64 is the number of gray scale levels one
has) and used as a mapping table.

Now, this is 1D only. Does anyone have an idea, wether anything
similar could be done in a 3D RGB color space? Any suggestions?
Jochen Schwarze                     Domain: schwarze@isaak.isa.de
ISA GmbH, Stuttgart, West Germany   UUCP:   schwarze@isaak.uucp
                                    BITNET: schwarze%isaak.uucp@unido.bitnet
                                    Bang:   ...!uunet!unido!isaak!schwarze