This is a multi-part message in MIME format.
------_=_NextPart_001_01C8B503.CC8C8DB0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Terje wrote:
> (It seems intuitively obvious that you must first determine the 3D =
outer
> boundary of the volume that surrounds all the true color pixels, then
> make sure that the surface described in the same space by your palette
> selection can cover the same volume: Any point outside the outside
> boundary is impossible to reach using any linear combination of the =
pal
> entries on the inside, right?)
Right. And as most palette reduction algorithms shrink
the number of colors by averaging a span of colors, the
gamut is reduced. With 256 colors this usually is not too
bad. But with 16 color images, the error diffusion can
cause some "amusing" results until you debug your code, as
large ****tions of the image are outside the reduced palette.
Implies that the palette reduction problem can change with
the final number of colors wanted. The pairwise clustering
algorithm that I use works nicely for 256 colors. And it
sometimes works well with 16. But it is slow.
Steve N.
------_=_NextPart_001_01C8B503.CC8C8DB0
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Diso-8859-1">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
6.5.7653.19">
<TITLE>Re: MMX speedup for Floyd Steinberg error diffusion</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<P><FONT SIZE=3D2>Terje wrote:<BR>
<BR>
> (It seems intuitively obvious that you must first determine the 3D =
outer<BR>
> boundary of the volume that surrounds all the true color pixels, =
then<BR>
> make sure that the surface described in the same space by your =
palette<BR>
> selection can cover the same volume: Any point outside the =
outside<BR>
> boundary is impossible to reach using any linear combination of the =
pal<BR>
> entries on the inside, right?)<BR>
<BR>
Right. And as most palette reduction algorithms =
shrink<BR>
the number of colors by averaging a span of colors, the<BR>
gamut is reduced. With 256 colors this usually is not too<BR>
bad. But with 16 color images, the error diffusion can<BR>
cause some "amusing" results until you debug your code, as<BR>
large ****tions of the image are outside the reduced palette.<BR>
Implies that the palette reduction problem can change with<BR>
the final number of colors wanted. The pairwise clustering<BR>
algorithm that I use works nicely for 256 colors. And it<BR>
sometimes works well with 16. But it is slow.<BR>
<BR>
Steve N.</FONT>
</P>
</BODY>
</HTML>
------_=_NextPart_001_01C8B503.CC8C8DB0--
|
