Received: from nova.unix.portal.com by netcom21.netcom.com (8.6.9/Netcom)
id QAA26956; Sun, 5 Feb 1995 16:13:01 -0800
From: DrGandalf@cup.portal.com
Received: from hobo.online.portal.com (hobo.online.portal.com [156.151.5.5]) by nova.unix.portal.com (8.6.9/8.6.5) with ESMTP id QAA05113 for <lightwave-l@netcom.com>; Sun, 5 Feb 1995 16:12:21 -0800
Received: (pccop@localhost) by hobo.online.portal.com (8.6.7/8.6.5) id QAA23933 for lightwave-l@netcom.com; Sun, 5 Feb 1995 16:12:20 -0800
To: lightwave-l@netcom.com
Subject: Re: Anti-aliasing problem.
Lines: 47
Date: Sun, 5 Feb 95 16:12:19 PST
Message-ID: <9502051612.3.22420@cup.portal.com>
X-Origin: The Portal System (TM)
Sender: owner-lightwave-l@netcom.com
Precedence: bulk
>I have created an object which is beige in colour, and which has thin
>dark (black) lines running horizontally along it. The object was
>rendered at something approaching an isometric view which makes
>this dark lines render at around 20-30 degrees on the screen.
>
>I have been noticing that LW is not rendering these lines properly.
>It's not just these black lines that are doing it - white highlights
>too - plus it's not just these 30 degree lines.
>
>The lines turn out something like this:
>
> ..:::II HHHHHHHHHHHHMMMMMMMMMMMM
> MMMMMMMMMMMMHHHHH IIII:::...
>
>It does that all the lenth of the line, so it looks like it's broken
>every cm or so. Either side of the gap IS actually correct, it's just
>a small break at a certain position all along it.
>
>I've just uploaded the object to tomahawk (Penti-Umm2.lha) although I
>did go over all the really noticeable line in Brilliance and fix them
>up. However, there are still parts which show this effect.
>
>So what is causing this?
>
This is caused by aliens. No, sorry, alias errors. A different sort of
aliasing error. (forgive me, I just got back from a gig, and my mind is
still recovering)
What happens is that part of the thin, slightly diagonal line is getting
lost between scan lines. So, when the first image is rendered (before the
"Searching for edges" step is done, according to your adaptive sampling
setting) the thin line would look something like this (forgive the ASCII):
__________ ----------
The space ^^^ here is "between the scanlines", and the line is missed
there. The adaptive sampling routine misses this spot, since there is
no edge to detect. It anti-aliases as you described above.
The solution? You can turn off adaptive sampling, that may help a bit. The
other solution is to render an oversized image (2X or even 4X) and then
shrink it. All solutions to this problem will cause very large increases in