Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!decwrl!sgi!shinobu!odin!tom.dallas.sgi.com!fouts
From: fouts@tom.dallas.sgi.com (Christopher L. Fouts)
Newsgroups: comp.sys.sgi
Subject: Re: Fooling with 'tops' and 'imgexp' utilities
Message-ID: <1991Apr11.225728.28562@odin.corp.sgi.com>
Date: 11 Apr 91 22:57:28 GMT
References: <1991Apr10.010515.19951@leland.Stanford.EDU>
Sender: news@odin.corp.sgi.com (Net News)
Reply-To: fouts@tom.dallas.sgi.com (Christopher L. Fouts)
Organization: Silicon Graphics, Inc. (Dallas)
Lines: 38

In article <1991Apr10.010515.19951@leland.Stanford.EDU>, dhinds@portia.Stanford.EDU (David Hinds) writes:
|> I'm curious about the parameters for the 'tops' program.  I remember reading
|> that the default halftone screens on Laserwriters were very nonoptimal; are
|> there some values that work better with 'tops', or are its defaults as good
|> as they can be?  Also, I can't seem to get a plain white background with my
|> printed images.  I used the 'invert' program to flip the screen black to
|> white, but it comes out as a sparse dot pattern on our laser printer.  So,
|> I tried using 'imgexp {files} 10 245' to force the very light and very dark
|> areas to true white and black.  When I look at the images before and after
|> with 'hist', however, they seem to have exactly the same grey scale ranges!
|> There are still lots of vertices with grey values in the 251..254 range.
|> 'imgexp' does seem to work if I ask for more extreme range reductions.

There is a slight bug (feature?) in the routine to convert RGB to black and
white.  In tops.c (as of 3.3.2), the routine rgbrowtobw has the following line:

            *obuf++ = (77*(*rbuf++) + 150*(*gbuf++) + 28*(*bbuf++))>>8;

Notice that for a gray RGB value (rbuf = gbuf = bbuf = value ), this is the
same as

	    *obuf = ( (77+150+28) * value ) >> 8 ;   or
		  = ( 255 * value ) / 256

Thus white RGB values (value = 255) get rounded off to 254, which results in
a very sparse dithering.  Correct this by adding 1 to the constant of your
choice, such that the total is 256, as in

            *obuf++ = (77*(*rbuf++) + 150*(*gbuf++) + 29*(*bbuf++))>>8;
						      ^^
 
-- 
Chris L. Fouts                           email: fouts@dallas.sgi.com
Systems Engineer                         vmail: 8714
SGI - Dallas                             phone: 214-788-4122

"To me, boxing is like a ballet except that there is no dancing, no
 choreography, and the dancers hit each other." - Saturday Night Live