Edge detection at specific orientation

Hi to all,
I have a general question about edge detection.
Someone knows some technique/algorithm to find only edges in an image with some specific orientation, not necessary with fiji?
I know in general how to obtain 45°, vertical, horizontal but I would like to know how to design a specific fiter to obtain for example edges more or less perpendicular to a line.

Thank you all,
emanuele martini

Hi @emartini

Horizontal edges in an image give a high response to a Difference of Gaussian filter where the filters are shifted vertical.

To detect a line which is rotated by 5° away from the horizontal line you could just rotate your DoG filter by the same amount.

As a rule of thumb: A kernel detects structures which look similar to the kernel.


Thank you I will try

You can also create your own Gabor filter for that specific orientation.


My group uses curvelets to do this, but all the code is currently in C and MATLAB. We unfortunately do not have any ImageJ plugins available yet. But perhaps the algorithms are still useful to you.

/cc @guneetsinghmehta


yes, curvelets might be a good option for this problem. In our tool, we simply select curverlets or large curvelet coefficients from curvelet transform to represent edge features. This approach avoids using any other edge detection algorithms and works pretty well under low SNR conditions, for instance, for the line-like collagen fiber alignment in our study.

1 Like