From: walter_at_[hidden]
Date: 2001-11-19 07:14:52
--- In boost_at_y <mailto:boost_at_y>..., Ullrich Koethe < u.koethe_at_c
<mailto:u.koethe_at_c>...> wrote:
> Karl MacMillan wrote:
[snip]
> > Before considering your library specifically, I would be
interested in what
> > people think about how image processing algorithms could
potentially relate
> > to the Matrix classes currently being worked on for boost.
>
> I think it would be good to have interoperability between the Matrix
> classes and image processing operations. However, when I last asked
the
> Matrix people they didn't seem to be interested in this problem. The
> design of a good matrix class was probably enough of a challenge in
> itself.
That's quite true.
> > Also, it seems
> > that it would be helpful to develop a set of multi-dimensional
iterator
> > categories analogous to the current STL categories to enable a
more formal
> > separation of the storage of the containers and algorithms.
>
> I have a formal specification for 2D iterators. The relevant papers
are
> included in the VIGRA documentation (also available at
> <http://kogs-www.informatik.uni-
hamburg.de/~koethe/papers/GenericProg2DC++Report.ps.gz>
> and
> <http://kogs-www.informatik.uni-
hamburg.de/~koethe/papers/handbook.ps.gz>).
> I could readily write a similar specification for the image data
> structures.
Amongst other things we've written a draft for 2D iterator concepts.
You can find it in the sourceforge CVS
at /boost/libs/numeric/ublas/doc/iterator.htm. I still believe
describing STL-like 2D iterators is a major challenge ;-).
> A more general question is whether iterators are necessary at all.
> Virtually all image data structures support random access, so
indexing
> would be sufficient. What's your opinion?
Iterators have an interesting performance behaviour on large data
sets (with most compilers I've seen so far).
Regards
Joerg