Newsgroups: comp.sys.atari.st
Path: utzoo!utgpu!watserv1!watmath!lsuc!jimomura
From: jimomura@lsuc.on.ca (Jim Omura)
Subject: Re: GemView (new version and some questions)
Organization: Consultant, Toronto
Date: Thu, 23 May 1991 09:06:38 GMT
Message-ID: <1991May23.090638.25503@lsuc.on.ca>
References: <3272@laura.UUCP> <1991May22.193554.26435@lsuc.on.ca>

In article <1991May22.193554.26435@lsuc.on.ca> jimomura@lsuc.on.ca (Jim Omura) writes:
>In article <3272@laura.UUCP> haacke@exunido.UUCP (Ralf Haacke) writes:

...

>>- A new color reduce algorithm. I have the followed idea:
>>  A color is a point in a 3-dim-room (N^3). First calculate the distance
>>  from each point to each other and save the nearest point with the distance.
>>  Next searching the nearest neighbours and make this two point to one. The
>>  new points is nearer to a  point which represent more points then the
>>  other. From the new point calculate the distance to all other points and
>>  save the nearest (some other points must update, if the new point is nearer
>>  as the nearest neighbour before). Then begin to search the nearest ....,
>>  until we have reduce enough.
>
>     I felt that you did a really good job on the colour reduction
>algorythm.  The colour selection seems better than on a couple of
>other programs I've been using ("VDI_GIF" and another one I can't
>remember offhand).  I've been kicking around ideas for colour
>reduction codes for a while now and I haven't gotten around to
>writing one yet, but it seemed to me that it might be easier to
>work in a 2D framework based on Hue and Intensity, which is the
>way a television signal works.  But to do this you have to start
>by converting the palette from RGB's for most picture format, then
>do the calculations, and then re-convert to the target palette
>style (Atari in this case).

     So there I was trying to sleep when it hit me at about 4:30 AM
that I was wrong.  This *still* isn't a 2D system because you still
have to consider "Saturation".  There doesn't seem to be any way
to reduce it to 2D.  At least I don't think so.  I still might
use Hue/Intensity/Saturation instead of RGB if I get around to
writing a conversion.  No rush though.  I have other things to
get to first.

>     Working from Hue and Intensity, I think you could give preference
>to one or the other for your colour assignments, as an option for
>the user.  That is to say you could allow the user to chose whether
>colour assignments are weighted to give the maximum detail in terms
>of the hues (which could result in a pastel, low contrast image with
>a lot of detail in terms of colours) or maximum detail in terms of
>contrast (which might result in a picture with fewer colours, but
>more shading).
>
>     Again, I'll emphasize that I haven't gotten around to writing
>this code yet so I don't know if it's going to work out the way
>I'm expecting it to.


-- 
Jim Omura, 2A King George's Drive, Toronto, (416) 652-3880
lsuc!jimomura
Byte Information eXchange: jimomura