Newsgroups: sci.image.processing
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!gatech!howland.reston.ans.net!news.sprintlink.net!mv!news.imagauto.com!not-for-mail
From: tut@imagauto.com
Subject: Re: image independent color quantization
Message-ID: <597cb$9191d.2de@news.imagauto.com>
Date: Tue, 09 May 1995 14:25:29 GMT
Reply-To: tut@imagauto.com
Organization: Imaging Automation
References: <3odfpc$sqb@ghost.sm.dsi.unimi.it> <AZ.95May8170740@saturn.analog.com>
X-Newsreader: IBM NewsReader/2 v1.09
Lines: 30

In <AZ.95May8170740@saturn.analog.com>, az@saturn.analog.com (Alex Zatsman) writes:
>
>In article <3odfpc$sqb@ghost.sm.dsi.unimi.it>
>attias@pippo.sm.dsi.unimi.it (Roberto Attias) writes:
>
>
>>    Hi. I would like to have some suggestion about color quantization.
>>    I want to use an image independent color quantization (fixed palette).
>>    Can you give me some hints? 
>
>Yes, here is one:  DON'T.
>If you do it with 8 bit colors, you either have to dither your image
>to approximate colors, or you'll have a significant color distortion. 
>The whole reason for palettes is that 8 bits is not enough for color!
>
>--
>Alex Zatsman, Analog Devices, Inc.,
>alex.zatsman@analog.com
>

This is most certainly incorrect.  On a display with suitable eye closure
angles, fixed palette techniques are quite useable.  In fact, at a viewing
distance of >= 18", you will be hard pressed to tell the difference from
24 bit color.

Bill Tuttle
Imaging Automation




