Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!cbatt!ihnp4!qantel!intelca!oliveb!hplabs!tektronix!teklds!zeus!bobr
From: bobr@zeus.UUCP (Robert Reed)
Newsgroups: net.graphics
Subject: Re: jaggie remover
Message-ID: <598@zeus.UUCP>
Date: Fri, 26-Sep-86 15:05:16 EDT
Article-I.D.: zeus.598
Posted: Fri Sep 26 15:05:16 1986
Date-Received: Tue, 30-Sep-86 07:18:27 EDT
References: <281@joevax.UUCP>
Reply-To: bobr@zeus.UUCP (Robert Reed)
Organization: CAE Systems Division, Tektronix Inc., Beaverton OR
Lines: 44

Steve Hawley in <281@joevax.UUCP> provides what he calls a "jaggie remover".
This is a very crude algorithm which I would not recommend to anyone
seriously interested in anti-aliasing.  Ignoring the blantant programming
errors in his sample, his technique is a blending algorithm which averages
the aliased samples around and including the pixel.  I think it is intended
to avoid smearing by staggaring the changes behind the leading edge of the
averager, but suffers the problems of filtering without oversampling.   This
algorithm will:

    o  Give believable effects on the edges of large solid polygons.

    o  Wash out thin lines and text

    o  Generally destroy textures.

The problem is that the equal weighted sample does NOT take into account the
underlying structure of the picture.  Consider what it will do to this image
of a horizontal line, where we have pixel with 4 levels of brightness.
(o = off, + = dim, * = semi-bright, # = bright)

before "jaggie removal":

    oooooooooooooooooooooooooooooooooooooooo
    oooooooooooooooooooooooooooooooooooooooo
    ########################################
    oooooooooooooooooooooooooooooooooooooooo
    oooooooooooooooooooooooooooooooooooooooo

after:

    oooooooooooooooooooooooooooooooooooooooo
    ++++++++++++++++++++++++++++++++++++++++
    ****************************************
    ++++++++++++++++++++++++++++++++++++++++
    oooooooooooooooooooooooooooooooooooooooo

The line has been expanded to three rows and dimmed.  

Once again I emphasize, this is NOT a topic for novice speculation.  Much
reseearch has been done on this topic, and if you are really interested,
consult one of a number of texts which have been mentioned in this
newsgroup.  Please do not inflict off-the-cuff schemes on this audience.

Robert Reed, Tektronix CAE Systems Division, bobr@zeus.TEK