Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uwm.edu!psuvax1!rutgers!att!cbnews!mjs
From: mjs@cbnews.ATT.COM (martin.j.shannon)
Newsgroups: comp.graphics
Subject: Re: Comp.graphics.images
Summary: 16-bit LZW can be a memory hog
Keywords: GIF87a, GIF90a?
Message-ID: <12083@cbnews.ATT.COM>
Date: 7 Dec 89 18:27:20 GMT
References: <2045@netxcom.DHL.COM>
Reply-To: mjs@cbnews.ATT.COM
Organization: AT&T Bell Laboratories
Lines: 39

In article <2045@netxcom.DHL.COM> jallen@netxdev.UUCP (John Allen) writes:
>The GIF standard uses a slightly modified 12 bit Lempel-Ziv algorithm
>to compress the image data.  Since compress is also Lempel-Ziv, the size
>of a GIF is about equivalent to a 12 bit compress.

Plus the header information, plus the pallette, neither of which is
compressed at all.

>So any gain obtained from compressing a GIF file will come
>from full 16 bit compression.

Certainly many images would gain from using 16-bit compression, but
doesn't that pose some serious memory usage problems for 8086-based
machines (and other small-address-space machines, too)?  I do all my
GIF viewing & manipulating on a '386 running SVR3.2, so moving up to
16-bit compress doesn't bother *me*, but is it a good idea to make it
so difficult for the small machines?

>Perhaps it would be worthwhile to lobby for an enhancement to the GIF
>standard which permits 16 bit LZW compression.  If GIF did use 16 bit
>compression, the result should be much smaller.

This is probably a very good idea, but there are a few other things
that really want to be added to the specification, as well.  Among them
are: a comment/copyright block (arbitrary text that *must* be displayed
when the picture is viewed); aspect ratio indicator; an RLE mode (many
of the pictures I've seen would compress much smaller if they were 1st
run through an RLE encoder); there are others.  I made a list about a
year ago, when I first figgered out how to read a GIF file, but I don't
have it handy -- I'll try to dig it up & post it.

>John Allen, NetExpress Communications, Inc.    usenet: jallen@netxcom.DHL.COM
>1953 Gallows Road, Suite 300                   phone:  (703) 749-2238
>Vienna, Virginia, 22182                        telex:  901 976

-- 
Marty Shannon; AT&T Bell Labs; Liberty Corner, NJ, USA
(Affiliation is given for identification only:
I don't speak for them; they don't speak for me.)