I have a binary image with uneven and straight lines that I want to characterise in means of its regularity. Now I'm doing it using the box-counting method, but I'm aware that maybe other procedures could be used. Which one do you recommend? Is the morphological dimension the most adequate for this situation?
dear Clara, does your image represent a 2D or a 3D structure?
During my PhD I used fractal dimension to characterise natural landscape features and rock masses in mines but I think that every geometrical analysis can be comparable.
In such a case, my experience is that the box-counting method is the most stable. Moreover, it is also capable of giving you a fractal dimension which is related to both the spatial distribution of objects (in the 2D or in the 3D space) and the frequency size distribution of the same objects.
To be able to help you further, however, I would need to have a look at one example of the binary images you are using.
In the meantime, good luck for your research!
best
Filippo
Box counting can be simple and effective. I mean if you are looking in one direction (+x only) for each line and if the neighbouring one (i+1) is not the same as the one you are in at that time, you can assign a number for it, and add them at the end of the algorithm, so you can say the image is something % (ir)regular. Hope that's what you want :) It will work for colored ones too.
Another thing comes to mind, which is looking to the picture in the frequency domain also.
- Badges
- Science topic