Re: [Ltilib-devel] lti-lib2 / CVR status

[Pablo A. <pal...@ie...>]

Hello,

well I am quite active in the development of the CVR-Lib 
(http://cvrlib.sourceforge.net)  which is just the next version of the 
LTI-Lib.  

The history is as follows.   1998 with my project colleagues at the 
RWTH-Aachen, Germany, we decided to create a library to join forces.  The 
LTI-Lib 1 was born.  2003 when I finished my thesis, the library grew quite 
impressively, so that I decided to spend some time trying to fix some flaws 
in the LTI-Lib 1 design, specially those related to dealing with the 
incompatibilities of MS Visual C++ 6 with the ANSI C++.  We removed very 
awful workarounds, and clean up the design a little bit and then the LTI-Lib 
2 was born.  Peter Doerfler and I were practicaly the only ones developing 
that library.  I kept working on it here in Costa Rica more under a on-demand 
approach, what we need get ported from the LTI-Lib 1 or is redesigned.

The name "LTI" comes from "Lehrstuhl für Technische Informatik" (Chair for 
Computer Engineering), and the sad part is that that Chair ceased to exist a 
few years ago, so that nobody there does anything with the library.  
Additionally, the LGPL turned to be somehow limiting for the use of the 
library in the industry.  So the CVR-Lib is just a few months there, using 
the more liberal BSD license.  This can help the CVR-Lib developers to get 
some money from the industry to develop, and the industry pays some attention 
to our project.  However, I'm the only one working on that by now, because I 
want to do a release as soon as the new core concepts are settled down.  The 
major differences are first, that in CVR-Lib the matrices are always 
connected.  I still have to design a submatrix class that replaces 
the "lined" mode of the LTI-Lib 1 and 2.  That mode was practically never 
used, and has been a head-ache in the implementation of algorithms that might 
get in trouble if the matrices were in fact submatrices...  Another major 
change is the convolution class, which has been the major flaw in the LTI-Lib 
concept, since it is too slow, and the code has been awfully big and 
unmaintainable.  I am splitting the class to cope with several cases that can 
be speeded up considerably.

I work now for the Costa Rica Institute of Technology, and we are using the 
library in several research projects, so I am keeping at least the CVR-Lib 
alive, with help of many of my students of course.  If you want to cooperate, 
that would be wonderful!  That's the idea of making our efforts public!

Kind regards,

Pablo


On Wednesday 20 February 2008 12:11, Per Rosengren wrote:
> I just read the developers guide, and it really impressed me! I too
> would like to contribute to this project, just as Tamaki Toro wrote. I
> discussed this with a colleague in my institution (CVAP, KTH in
> Stockholm), and he is also interested in joining forces to create a
> replacement for opencv.
>
> What is the status on ltilib2 or CVR. Pablo Alvarado, are you working on
> it? Tamaki had heard something about it, and I have seen it in CVS, but
> I don't find any information about it. Which one should I use?
>
> Who is working with what?
>
> How complete and bug free is lti-lib and CVR?
>
> Why isn't lti-lib in the debian repositories? I found this wishlist:
> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=449044

-- 
Dr. Pablo Alvarado			E-Mail:	pal...@ie...
Escuela de Ingeniería Electrónica	Tel.:	(+506) 550 2106
Instituto Tecnológico de Costa Rica	Fax.:	(+506) 591 6629
Apartado Postal 159-7050
Cartago
Costa Rica