home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Amiga Plus 1995 #5 & #6
/
Amiga Plus CD - 1995 - No. 5 and 6.iso
/
pd
/
grafik
/
lightwave
/
lightwave-apr95
/
000677_owner-lightwave-l _Mon Apr 24 15:36:29 1995.msg
< prev
next >
Wrap
Internet Message Format
|
1995-05-06
|
2KB
Return-Path: <owner-lightwave-l>
Received: by netcom6.netcom.com (8.6.12/Netcom)
id KAA24477; Mon, 24 Apr 1995 10:58:45 -0700
Received: from gaspra.pd.com by netcom6.netcom.com (8.6.12/Netcom)
id KAA24452; Mon, 24 Apr 1995 10:58:37 -0700
Received: by gaspra.pd.com (8.6.9/1.37)
id KAA10651; Mon, 24 Apr 1995 10:58:48 -0700
Date: Mon, 24 Apr 1995 10:58:47 -0700 (MST)
From: Ernie Wright <ernie@gaspra.pd.com>
To: lightwave-l@netcom.com
Subject: RE: YET ANOTHER SUGGESTION N
In-Reply-To: <9504240642.AA007dz@fusion.mv.com>
Message-ID: <Pine.SUN.3.91.950424103644.10498A-100000@gaspra.pd.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-lightwave-l@netcom.com
Precedence: bulk
Alan Chan:
> Is the algorithm patented? How does Ultimatte work?
Mark Thompson:
> I believe it is patented but I'm not certain. I have on several occasions
> asked around for ANY info or papers on the topic of automatic mask
> extraction (such as chroma keying) but have come up with nothing. Doing
> it well seems to be a well guarded trade secret.
If it's patented, the method is fully disclosed in the patent, which
anyone can read if they're willing to do the search. The library would
be a good place to look for papers.
Alan Chan:
> Would it be possible to look at the blue values from the RGB files and
> turn it into an 8-bit transparency?
Mark Thompson:
> Its actually a much more complex problem than it seems and this simple
> approach will not do it effectively. Consider that your blue background
> is neither a pure full intensity blue, nor is it perfectly uniform. Yet
> you want it to matte out completely while still preserving things like
> shadows, smoke, fine detail, or translucent surfaces. Pretty tricky stuff.
It doesn't strike me as being that hard. As a first try, translate the
RGB into HSV (hue, saturation, value) and tag the pixels lying near some
(bluish) H. If that doesn't work very well, try YUV or some other color
space. There's likely to be SOME linear transformation that makes the
matte color stick out like a sore thumb.
I suspect this isn't discussed in the context of digital graphics much
because it's only a problem when live video is involved, and the analog
approach for video has been working fine for three decades.
- Ernie