[curvelab-list] Hello, need some help

[rbaena at am.ub.es]

Hello,

  I am Roberto Baena, pHD student at the University of Barcelona, Spain.

  My thesis is about synthesis imaging in astronomy. A few weeks ago, I
discovered the curvelet transform and believed its application to the
images I am working with could be very interesting and fruitful. So I
decided to download the CurveLab software and installed in my laptop.

  Now, I am trying to understand the concepts and also trying to apply the
curvelets to my problem. I read the paper "Fast discret Curvelets
transforms", but there are many things that are still confused for me,
many of them avoid me to use the CurveLab soft in practice.

  Excuse me if the following questions are a little trivials but I prefer
to ask before going on with the work with the concepts/theory unclear.
Now I am using the wrapping curvelet transform.

     1º) When I obtain the coefficients 'c' from an image, I am supposed
to access them by c[s][w](i,j), where 's' is the scale, 'w' the wedge
and I guess (i,j) is the position or sample in that wedge.
         Suppose that I would like to create a new image with all the
curves detected. How can I do this? I mean, Do I need to create
's' images, one for each scale?

     2º) To create that image from the coefficients 'c', do I only have to
arrange them in a 2-D matrix?
         I mean, if I have an original image Im, that it is a matrix of
pixels of dimension N1xN2, from where I obtain coefficients
c[s][w](i,j). For obtaining a curve image, is it enough to put in
order the coefficients in a 2-D matrix N1xN2 (or 's' 2-D matrix)?
or are necessary intermediate operations?

     3º) How can I decide the width of the curves to be detected? Is this
the scale?


Well, I apologize again because I am sure the questions seem sillies, but
believe me, from outside and facing the problem for the first time
everything is a little confusing. So I prefer going step by step.

Thank you for your help and looking forward to hearing from somebody :-)

Roberto.