Eliminating artifact peaks in OrientationJ angle distributions

[smrithika]

I have been using the OrientationJ plugin to extract angular distribution of slightly noisy microtubule network images, and always obtained strong peaks at 0°. To counter-test this, I used a synthetic image of vertical and diagonal stripes (attached) and still obtain this ‘artificial’ peak regardless, as is confirmed by the color map that shows peaks of 0° indicated in turquoise(also attached). This is what is adding up in my filament images to give peaks at 0° that actually don’t exist.Can anyone explain why/how this happens and how to eliminate this problem? Thanks!

[gcardone (Giovanni Cardone)]

@smrithika The artifacts come from the shape of the objects in your synthetic example. Since they are rectangles, their small side accounts for the peaks at 0 and -58 degrees, complementary to the two dominant directions. While this explains the results for the synthetic image, I am not sure this would also explain the artifacts in your experimental images. You would need to upload a small example to get additional feedback. Just keep in mind that different approaches will provide different results, depending on the algorithm used. For example, if you try the Directionality (Fourier components option), available by default in Fiji, these artifacts are much less pronounced.
I hope it helps.