Hi, Lan,
Thank you very much for your quick and informative response.
LAPACK, BLAS, and MINPACK are all Fortran based, am I right?
Should I install those Fortran compiled libraries separately before installing or building VXL?
Is there any website comparing these libraries? I am more interested in the coverage of their functionalities. For example, LAPACK and BLAS are more linear algebra (vector, matrix computation), etc.
We probably need a table if not yet to give a summary about these libraries.
Thanks again for the information you provided about VXL.
Shoujie He
On Tue, Jul 8, 2008 at 10:22 AM, Ian Scott <ian.m.scott@student.manchester.ac.uk> wrote:

VXL includes a large section of LAPACK, and BLAS as the back-end to our linear algebra code. It also uses minimisers (MINPACK), and other functionality from the netlib site.

When we last compared VXL and openCV (many many years ago) we decided to go with VXL because of its strong C++ orientation and more consistent coverage of the basic concepts we used. OpenCV is mostly C and had a wider range of top class existing Computer Vision in it, more support for real time camera acquisition.

If you are trying to decide with library to use, then I would suggest going mainly with one coherent library- Raw C/LAPACK etc, or openCV, or VXL. A few others that might interest you are Matlab, ITK, or the Boost libraries. (The latter acquired an imaging library a few months ago.)

Pick one because it is more efficient to code consistently within the paradigm that each library provides. You can always interface between the libraries, if need be, for efficiency or to pick up an extra algorithm or two.

As a means of picking one, I would suggest sending an email to each group outlining the work you want to do and listing your requirements, and asking what relevant code/examples/documents already exists. The replies should help you decide where to go.

If you have already decided to use VXL, then welcome. But I'm not sure how to answer your question usefully, but here goes: We use the core of VXL a lot, and occasionally do visualisation or prototyping work in Matlab or ITK/VTK. We have our own private library that is built on top of VXL to run our own experiments (Statistical shape models, AAMs, volume registration, etc.) Occasionally we integrate small 3d party libraries into our code. The simpler/more robust/more generally useful stuff we publish in VXL.

If you have a more specific question, then please do ask.


Shoujie He wrote:
Hello, everyone,
 I am new to VXL and reading the online VXL book.
It seems that some of the functionalities in VXL duplicates with
boost, OpenCV, LAPACK, etc. I wonder if any experience user of VXL could share with me your experience about the best way of using VXL. I wonder if VXL covers all the functionalities provided by LAPACK (CLAPACK).
 Thanks in advance.
 Shoujie He