On Wed, 22 Aug 2007 23:19:30 +0100, David Marrs <[EMAIL PROTECTED]> wrote:
> Amit Kumar Saha wrote:
> > Am i thinking in a way that could possibly be implemented? or is the
> > word "extensible" remotely applicable to my idea?
> > 
> So what you're talking about here is a graphical interface to the API that a 
> user can use to build his own extensions? Essentially, it's a graphical 
> programming language. There are quite a few examples of these in the music 
> world 
> that allow the construction of modular synths (amongst other things).

Not only in the music world.  Around 1993 or so, I remember using a
program called Khoros Cantata to design some image filters.  A quick
search on the web returned this tutorial with screenshots of the old
X11 version that I used:
Just imagine that the glyphs (the boxes that can be connected) could
be the operations that GIMP offers through the PDB.

It seems that Khoros Cantata has been acquired at least twice by
different companies, has been ported to Windows, and is now sold as
Accusoft VisiQuest:
The interface has changed a bit, but it is still based on the same
ideas of connecting glyphs providing various functions.

By the way, this brings back some memories about what I did in 1993
as an exercise for a course in digital imaging.  Students were
asked to write a new filter for Khoros and use it to perform some
interesting image transformations.  I decided to write a filter
that would use various correlation methods to reconstruct an image
in which the lines were shifted horizontally by some random amount.
The default method was optimized for "un-shifting" images at TV
resolution (PAL) that were shifted by one of three predefined
offsets picked at random.  The method worked surprisingly well.  The
code and the results were presented to the professor.  He asked if,
by any chance, this could be used to decode Canal+ images (a pay TV
channel that was popular in several European countries).  And the
answer was yes.  Not in real time because the 486 processors that
were common at that time were not fast enough, but the method worked
well with recorded TV streams (sometimes even without using inter-
frame correlations).  Ah, I still have fun when I remember this and
when I consider that the professor gave one of the top marks for
this exercise.

Gimp-developer mailing list

Reply via email to