|
Boost : |
From: brianjparker_at_[hidden]
Date: 2001-11-15 15:59:40
--- In boost_at_y..., "Peter Dimov" <pdimov_at_m...> wrote:
> From: "Ullrich Koethe" <u.koethe_at_c...>
> > BTW, what's your image library doing? Is it also public?
>
> To reply to the specific question:
>
> Currently I focus on operations relevant to computer games. These
include
>
> * import/export (you've got this covered)
>
> * resampling
>
> * copy (as fast as possible) image A onto image B at point (x, y)
using
> effect E (additive blend, alpha blend, multiplicative blend... like
the
> photoshop layer effects)
>
> * (quickly) convert image A to the specified packed-pixel format,
like 555
> RGB.
>
> * crop image A leaving only the parts where the alpha channel is
not zero.
Hi,
As a game developer, I presume you are targeting x86 platforms.
If so then you can't really go past the (free, but not open source)
Intel Performance Primitives library-
http://www.intel.com/software/products/perflib/
More generally, I think that such low-level image processing
operations are best handled by such a set of optimised standard
functions (much like the BLAS is used for linear-algebraic
primitives) rather than duplicating this code for each array class.
The IPP has an excellent low-level design and so would be an ideal
candidate for such a standard low-level library, IMHO. Various array
classes, (VIGRA, valarray etc.) can then have wrappers written to
provide convenient access to this functional interface.
Another excellent site for more advanced machine vision algorithms is
the (open-source) OpenCV library at
http://www.intel.com/research/mrl/research/opencv/.
,Brian Parker
Senior Research Engineer,
Canon Information Systems Research, Australia
PhD student,
Biomedical & Multimedia Information Technology (BMIT) Group,
Computer Science Department, Sydney University, Australia
(brianjparker_at_[hidden])
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk