Orientation J and its interpretation for electrospun nanofibers

Hello all,

I have electrospun nanofibers that I want to calculate the align of nanofibers. I did with orientation j which gives me 3 values energy, orientation and coherency. But I don’t know how can exactly use the orientation j and interpret the results. I read the papers which image j suggested.

Thanks alot


the following papers describe an alternative approach to lobal/regional orientation analysis:


If you manage to make the posted image visible to the forum, I shall try to apply this approach to your image and post the result.



Dear George

Thank you so much for your help. I copy the image again. Please see it.

Best Regards

Well Mina,

here is the windowed image that I used for the orientation analysis:

and here is the resulting graph:

Please note that in the graph, orientation angle begins horizontally (0 deg) and proceeds counter-clockwise. For the example image, the prominent orientation is observed at about 60 deg (red line) and the opposite at about 175 deg (blue line).

Orientation salience means that the analysis isn’t purely geometric but also depends to a certain degree on the contrast of the structures (for details see the first paper).



Dear George,

Thank you for your reply and help. Would you please say how obtain the graph? I have 6 other pictures which I should analyse.

Thanks alot again
Best regards

Dear Mina,

the processing is described in detail in the previously mentioned papers. For the shown graph I used the “Power Spectrum”-approach. (For your example image, the three possible approaches lead to essentially the same result.)

If you upload your remaining images somewhere I shall process them for you.



Thank you George. I mean that can I do same with Image j? I try my self then if I have question i will ask you.

Thanks alot

Dear Mina,

if you are an experienced ImageJ macro-coder, then you may succeed, but please note that the processing is non-trivial.



you find the ImageJ-plugin “Slice Integrals” that may help with the processing. However, it solves only part of the whole processing that must start with a non-trivial pre-processing step…

Good luck


Welcome to the forum, @mina and @hglu!

The original tiff file that was linked in the first post of this discussion is actually accessible but just not displayed, for technical reasons (most browsers don’t support showing tif format natively).
You can download the image using right-click and the context menu. Anyways, thanks @mina for posting it again in jpg format.

To mention yet another (and maybe easier) alternative: the Directionality plugin (Analyze > Directionality in Fiji’s menu) also produces a histogram of the most prominent angles present in the image.

For the theory behind directionality measurements, the mentioned literature is surely helpful.
In particular, for OrientationJ and the explanation of the terms coherency and energy, have a look at the PDF document linked from the OrientationJ page.

1 Like

Dear Jan,

actually the download doesn’t happen (Firefox for Mac). Of course I tried right-clicking.

Concerning “the Directionality plugin” I should like to mention that the results shown above are much more accurate. Furthermore, the approach of much respected Monsieur Tinevez doesn’t take into account the angular sampling theorem.

OrientationJ uses a completely different approach: It works locally and there is no influence of the contrast of image structures.

Anyway, everybody is free to use the approach she likes most but I suggest to understand the math before deciding…

All the every best


Oh, that’s strange. It works for me with the same combination (Firefox 50.1.0 on Mac OSX 10.12.2). Maybe there is a difference in the installed add-ons?

Sorry, I’m not that familiar with the theory here. I just wanted to mention the existence of the other plugin as well. From what I can see, Directionality offers the choice of two different methods, based on Fourier components, or on local gradient information.

Absolutely agreed.

Have a nice weekend.


I analysed two different samples of electrospun nanofibers by orientation J. I want to compare the alignment and orientationf samples that which of them has higher orientation. How can I interpret? What the Y axis show us? I mean is Y axis the number of nanofibers placed in special angle?

Thanks a lot

Well Jan,

maybe you use a different Firefox configuration, although my version is 50.1.0 as well. My macOS is 10.11.6 however.

Directionality offers the choice of two different methods

Yes, but evidently the result you’ve shown is based on Fourier-analysis. The local analysis however, is rather related to what OrientationJ is able to provide.

The results from global and local orientation analyses differ considerably and the reasons for this are diverse. One of them I’ve stated above.

Let’s assume that the example image of Mina is sampled according to the sampling theorem, then we need—for global analysis—at least 678 angles to correctly display the information. If I do so using Directionality, I’m no quite happy with the result…

Let’s stick with math!

Have a nice weekend as well


Dear Mina,

can’t help you with OrientationJ, but if you compare the results from my analyses of the images, it is evident that Image-2 has a more pronounced and a higher maximum and less fluctuation elsewhere than Image-1.



which of them has higher orientation

I have no idea what you mean by “higher orientation”.

Image-2 has an angular dominance at about 152±30 deg. Apart from this orientation range the fibers appear more or less equally distribute. Not so with Image-1.



That’s very nice of you. I mean that to find that which sample formed in one direction. I tried to read the paper you sent me earlier. I found that analysed with different method and compared. as you suggested me I google “Power spectrum” but I don’t found any link to download to analyse other samples.

Thanks a lot

Sorry Mina,

but with “Power Spectrum” alone you are way off the track…
Actually, I didn’t suggest to “google ‘Power spectrum’” which actually will not help much with the problem in question!

In short, the first cited report compares three global approaches to orientation analysis (quality of the results & computational effort) when implemented digitally. Theoretically the three approaches lead to the same result which is shown mathematically in the second cited publication.

For your kind of images, the three digital implementations lead to results that are nearly indistinguishable, i.e. one may use the easiest approach which is the one based on the Power Spectrum. However, the most efficient approach, i.e. the one for which the relation of quality and relative computational effort is maximized, is the one based on the Autocorrelation function.

As mentioned before, the implementation of the three approaches is not trivial and surely not a project for a beginner. The provided ImageJ-plugin “Slice Integrals” may help but the remaining tasks, that are detailed in the report, are still quite involved. Actually, it took me some month to code, test and refine the three implementations that I cannot provide for free. However, I still offer to analyze a decent amount of images for you.

I mean that to find that which sample formed in one direction.

If you are looking for the sample having the greatest amount of aligned fibers (fibers having approximately the same orientation), then you must look for orientation results showing a single, high and slim (pronounced) maximum. In this respect, example Image-2 is better when compared to Image-1.



Dear Mia,

The FORUM told me that I’m not allowed to post any further replies, and that I’m limited to post a maximum of 5 images. What a mess!

Consequently, I propose to change to the ImageJ-list where we could continue our work.

For the list please see:

Please be so kind and post a short message there and I shall send you the results.



It’s nice of you. So I attached four more figure. To give a references for the analysis (to show how it is calculate) should I refer to the paper that you set me or any thing else?

Thanks alot

@MIna You may find this paper “Image analysis for measuring rod network properties” by Dongjae Kim, Jungkyu Choi and Jaewook Nam


to be very relevant to your work. They describe the method they developed (it seems to me it was done mostly in MATLAB), but have not released the code. I have already suggested that the authors consider implementing it in ImageJ/Fiji; they may consider doing this.

1 Like