Fragmentation Index

[Ion Andronache]

[Gabriel Landini]

On Saturday 02 May 2015 10:55:33 Ion C. Andronache wrote:
> interested in the degree of fragmentation of black pixels - foreground). It
> is possible to implement the formula
> http://ec.europa.eu/agriculture/publi/landscape/ch5.htm (5.3.2.Calculation
> of fragmentation index) or exist a method of calculation more specific? I
> look forward your advice. Best wishes,

That fragmentation index seems to rely on occupancy of the set under an
arbitrary grid. But what happens if you change the grid size? That leads to a
problem of scale: you change the grid size and the index changes so the
measurements are only accurate to a particular scale.

I think that a better approach would be to do a (fractal) box counting
analysis where each point in the log-log plot is the number of occupied grids
at various grid sizes and the fractal dimension (the slope, if the plot is
straight) explains how that occupancy rate changes with a change of grid size.
You can apply box counting with ImageJ. If you want to read a bit more about
that, I wrote an invited technical review for JMicr that might be useful and
has some examples. In your case the pattern seems to me a "mass fractal".
http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2818.2010.03454.x/full

cheers

Gabriel

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html