polite request - FracLac

[Matěj Lövy]

Hi all,

I have found ImageJ and FracLac plugin to estimate fractal dimension of
images. Although I am not a cell biologist but a student of zoology, I
consider the FracLAc ideal for calculating fractal dimension of planar
objects. We are analysing digitalised maps of burrow systems of subterranean
mammals.To describe complexity of burrow systems we used fractal dimension.
Till now, all analyses were done "handy" (especially box-cpunting method is
extremely time-consuming). Because of such reasons I tried to calculate
frac. dimension by FracLac. However, I face the problem how to get results
similar to those obtained by "handy technique". In "handy" technique the
grid calibres are derived from the longest side of an image analysed that is
256 mm. 256 mm is also length of the largest box, all smaller calibres are
calculated as 256/2n, where n is the exponent and takes values 1, 2, 3...8.
Then, Fractal dimension is calculated as the slope of ln N(e) against ln
(1/e), with the fitted line constrained to pass through the origin.
My problem is, that I do not know how to set the box-counting method in
FracLac to get similar result. If anybody could help, I really appreciate
all advices in this issue!!!

Many thanks in advance

best wishes
Matej Lovy


--
Matěj Lövy
University of South Bohemia
Faculty of Science
Branišovská 31
České Budějovice
370 05
Czech Republic
........................

phone: [phone]
email: [hidden email]

[Gabriel Landini]

On Wednesday 15 December 2010 13:19:15 you wrote:

> extremely time-consuming). Because of such reasons I tried to calculate
> frac. dimension by FracLac. However, I face the problem how to get results
> similar to those obtained by "handy technique". In "handy" technique the
> grid calibres are derived from the longest side of an image analysed that
> is 256 mm. 256 mm is also length of the largest box, all smaller calibres
> are calculated as 256/2n, where n is the exponent and takes values 1, 2,
> 3...8. Then, Fractal dimension is calculated as the slope of ln N(e)
> against ln (1/e), with the fitted line constrained to pass through the
> origin. My problem is, that I do not know how to set the box-counting
> method in FracLac to get similar result. If anybody could help, I really
> appreciate all advices in this issue!!!

Hi Matej,
There is a box counting built in ImageJ and you can specify the boxes:
 Analyze>Tools>Fractal Box count. This is quite fast.

I would suggest to use box sizes that are meaningful to the structure/pattern
that you are analysing, rather than the image size.

Why constraining the fitting line to pass through the origin? That does not
make sense unless you can prove that the scaling is present even at the
smallest scales (quite tricky to do when using digitised images).

Regards

Gabriel