Feature request: 3D Directional filtering

[andrewkennard]

I am a big fan of MorphoLibJ, especially the deep support for 3D image operations. I would love to have a 3D version of the Directional Filtering tool to allow me to enhance filaments in 3D stacks. Right now as I understand it (based on the manual) Directional Filtering is only for planar images, so it is applied slice-by-slice to 3D stacks. My filaments are largely in the Z direction.

Since the number of lines to test will get quite large in 3D, it might also help to include a "allowed angle" parameter that would generate linear structuring elements in the Z direction within a specific set of angles relative to the Z-axis.

I am lousy at Java programming otherwise I might offer to help!

[dlegland]

Hi Andrew,

thank you for your interest in MorpholibJ!

Yes, the 3D version of directional filtering would be great... You are right, for now it is only 2D. To develop a 3D version, one difficulty is to generate 3D orientations with an "appropriate" distribution -> try not to favor some specific directions, and to cover the unit sphere rather uniformly. This should be possible, but requires to be careful in the implementation.

I can try to have a look during summer. If you have some dataset that I could use as test images, that should be great, because I can make sure the method works...

[andrewkennard]

Thanks so much for the quick response! That does sound challenging. If you happen to have the time I would be very appreciative!

I can share a dataset. I think it is a good candidate for this approach because the linear structures in the stack are clearly visible by eye but their intensity is heterogeneous (sort of dots along a line) so I'm hoping that the directional filter will emphasize the long-range continuity. Here is a Google Drive link, let me know if you encounter issues with access: https://drive.google.com/drive/folders/1nZo2vL-QhxsWOmd6g1BRSWq2ohJaYoYP?usp=drive_link

In case it is helpful, I did some reading on the topic and came across this paper that seemed reasonably useful, and explicitly addressed the sampling issue that you mentioned: https://www.ias-iss.org/ojs/IAS/article/view/41-55

Thanks again!