Genuine Fractals - Yet Another Test/Opinion

["Eric Anderson" <eric@2live4.com>]

<x-charset iso-8859-1>Date: Fri, 10 Aug 2001 17:25:12 +0100
From: "Rew, Alan" <arew@gresham-computing.com>
Subject: RE: Testing Genuine Fractals - proof it does work

Richard (and the list):

I have extensively tested GF with photographic images.  The application in
this case was taking "low res" digital camera images sent over wireless
modems direct from the camera, and getting decent prints from them.  I did
this study while I was CTO at FlashPoint Technology - where we were working
on a wireless camera network before the tech "bust" sucked the money out of
the market.

In my tests, I showed that after you blow up the data by more than 4x (area)
or 2x in each direction, you begin to get the benefits from GF.  So, if you
are sending a QVGA over the wireless web and want a decent 3x5 or 4x6 print,
GF helps a lot.

How do you calculate this for a print?  Well, I think Epson uses 360 dpi as
their "base" when printing - i.e., they resize and sharpen images from
whatever the input dpi is to 360 before converting to ink blobs (OK, dots...
:)

So, if your blow-up is going to make your dpi less than 180 dpi, you will
get benefit from GF.  I am not saying that there is NO benefit prior to this
point, but it is hard to see.  As you blow it up more and more, the benefits
grow larger.

Regards,

Eric Anderson
Anderson Creations
http://andersoncreations.net

On Fri Aug 10, 2001  1:20 pm,
"Richard Wolfson" <rwolfson@LyricDesign.com>, wrote:

> Thanks, Royce, for the interesting report on Genuine Fractals. As usual
> from you, it's a contribution to this community. And as usual with GF,
> it answers some questions, but raises others.
>
> Here's the issue I see. For your test, you used "line art" -- drawings
> with smooth, hard edges. For this art, GF did better scaling up than
> Photoshop's bicubic algorithm. But I don't think this means GF would do
> any better than PS-bicubic scaling up continuous-tone art --
> photographic or fine art images -- which lack such smooth, high-contrast
> edges.
>
> What do you say? How about a more realistic test with continuous-tone
> images? For now, I think the test you posted falls short of "proof it
> does work" for that broader range of images.

-
Turn off HTML mail features. Keep quoted material short. Use accurate
subject lines. http://www.leben.com/lists for list instructions.

</x-charset>