Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

3D Tile Alignment #1

Open
tibuch opened this issue Mar 30, 2022 · 5 comments
Open

3D Tile Alignment #1

tibuch opened this issue Mar 30, 2022 · 5 comments
Labels
enhancement New feature or request

Comments

@tibuch
Copy link
Collaborator

tibuch commented Mar 30, 2022

Add 3D tile alignment script.

@tibuch tibuch added the enhancement New feature or request label Mar 30, 2022
@aloejhb
Copy link
Contributor

aloejhb commented Jul 6, 2022

Since the full 3D alignment pipeline is still on its way, could we write a script to roughly align the sections and put them into a volumetric file for exploring? This will help with data quality inspection and rough region of interest selection.

@aloejhb
Copy link
Contributor

aloejhb commented Jul 22, 2022

I added a minimum working example of rendering a roughly aligned multiscale 3D volume with TensorStore.

@aloejhb
Copy link
Contributor

aloejhb commented Jul 26, 2022

For roughly aligning sections, this script scripts/run_roughly_align_sections.py on the bo-dev branch can be a starting example.
There is an issue though. I crop a small region in the center of each section for computing cross-correlation, with size specified here.
However, for large shifts caused by changes in sections size, e.g adding a new row of tiles, the crop_size can be too small. It would be nice to have a smart way, to first try with the small crop size, and if the offset is too large, increase the crop_size until a proper offset is found.
In that way, most sections can be aligned with small cropped region with fast speed, and a few sections with large cropped region to accommodate big shifts.

aloejhb added a commit that referenced this issue Aug 7, 2022
@ganctom ganctom added this to the Selected for developement milestone Jan 10, 2023
@ganctom
Copy link
Collaborator

ganctom commented Jul 18, 2023

The registration procedure with diffident crop sizes is a good approach, however due to the imaging issues and deformation of stitched tiles the final shift vector should be computed from shift vectors computed from multiple regions on the section pairs. The precision of rough alignment will be better.

@tibuch
Copy link
Collaborator Author

tibuch commented Jul 24, 2023

The following should work:

  1. Compute coarse alignment (via the down-scaling approach) between section pairs. This can run in parallel.
  2. Aggregate transformations to obtain transformation w.r.t. the first section.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

3 participants