EnigmA Amiga Run 1997 July/EnigmA AMIGA RUN 20 (1997)(G.R. Edizioni)(IT)[!][issue 1997-07 & 08][EAR-CD IV].iso/lightwave/lwmlist/94.lightwave-09/000437_owner-lightwave-l _Thu Sep 15 09:41:39 1994.msg

Internet Message Format | 1994-10-01 | 3KB

Return-Path: <owner-lightwave-l>
Received: by mail2.netcom.com (8.6.9/Netcom)     id IAA17971; Thu, 15 Sep 1994 08:42:20 -0700
Received: from westford.ccur.com by mail2.netcom.com (8.6.9/Netcom)     id IAA16949; Thu, 15 Sep 1994 08:35:34 -0700
Received: from hubbub by masscomp.westford.ccur.com via TCP/IP with SMTP           id aa16699; 15 Sep 94 11:26 EDT
Received: from toaster by hubbub.westford.ccur.com           via TCP/IP with SMTP (local) id aa07350; 15 Sep 94 11:25 EDT
To: [e-mail]
cc: [e-mail]
Subject: Re: Antialiasing
In-reply-to: Your message of Wed, 14 Sep 94 21:25:59 -0700.              <[e-mail]>
Date: Thu, 15 Sep 94 11:25:37 EDT
From: Mark Thompson <[e-mail]>
Message-ID:  <[e-mail]>
Sender: [e-mail]
Precedence: bulk
Reply-To: [e-mail]

Ernie Wright <[e-mail]> writes:
> That's certainly more accurate, Mark, but for many folks I'm not sure
> it's clearer.

Perhaps your right. It is easy to confuse spatial sampling with edge checking.
Antialiasing is a very tricky subject.

> When adaptive is ON, edges are checked 5, 9 or 17 times.
> The reason this doesn't always work is that edge detection is based on
> the image pixels.

Exactly. Even though the edge detection makes multiple passes, the effort
is kind of wasted because detecting new edges on subsequent passes is pretty
unlikely when the only part of the image that is updated is the part that
detected edges in the first place. So if a sub-pixel edge is not detected
on the first pass and is not surrounded by other detected edges, it will not
be detected on subsequent passes.

> Stuart's explanation is the clearest (and most difficult).

I have to agree. He even helped clear up my understanding of motion blur/DOF
with adaptive enabled.

> But now I'm wondering whether the threshold test is based on the green
> level, or on something more like luminence, which is what Stuart hinted at.

I am pretty certain that the threshold test is based on intensity differences
for all three color channels (red, green, and blue) with the greatest edge
sensitivity on green. I forget the exact ratios (I believe John Gross posted
them once) but they are something like:
green theshold = threshold value
red theshold   = 2x threshold value
blue theshold  = 3x threshold value
%~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~%
%      `       '        Mark Thompson                 CONCURRENT COMPUTER  %
% --==* RADIANT *==--   [e-mail]        Principal Graphics   %
%      ' Image `        ...!uunet!masscomp!mark       Hardware Architect   %
%     Productions       [phone] [phone]   & General Nuisance   %
%                                                                          %
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~