abria / TeraStitcher

A tool for fast automatic 3D-stitching of teravoxel-sized microscopy images
http://abria.github.io/TeraStitcher/
Other
78 stars 32 forks source link

Stitched images with obvious seams and #32

Open Lalga opened 6 years ago

Lalga commented 6 years ago

Hey, So I tried using Terastitcher to stitch some image tiles I obtained, and the end result ending up looking like below. For the most part, the stitching worked well and the tiles were merged correctly, but there are some obvious errors as well. There are many sharp edges at the ends of tiles, suggesting that this tile likely isn't supposed to belong in that spot. This is most evident near the top of the image, and near the right side, where a vessel seems to be cut off abruptly. In addition, there is the large black space near the right and bottom edges of the image, where data should be. The raw data is definitely not blank, as I've done the same stitching with ImageJ and didn't run into this issue. Either way, any idea how to fix this? I've attached the xml file below.

capture Mouse hemisphere parameters.txt

iannellog commented 6 years ago

TeraStitcher proved to be quite robust in these years, and the problem you mention has not a clear reason. It is likely there is something wrong in the way you have used the tool, but of course this is not immediate to find. Could you please answer a few question just to clarify some point and exclude some possible problems?

-- Giulio

2018-07-06 19:04 GMT+02:00 Lalga notifications@github.com:

Hey, So I tried using Terastitcher to stitch some image tiles I obtained, and the end result ending up looking like below. For the most part, the stitching worked well and the tiles were merged correctly, but there are some obvious errors as well. There are many sharp edges at the ends of tiles, suggesting that this tile likely isn't supposed to belong in that spot. This is most evident near the top of the image, and near the right side, where a vessel seems to be cut off abruptly. In addition, there is the large black space near the right and bottom edges of the image, where data should be. The raw data is definitely not blank, as I've done the same stitching with ImageJ and didn't run into this issue. Either way, any idea how to fix this? I've attached the xml file below.

[image: capture] https://user-images.githubusercontent.com/33734415/42391052-feaadbee-811b-11e8-8bc4-cbb783edb2a3.PNG Mouse hemisphere parameters.txt https://github.com/abria/TeraStitcher/files/2171365/Mouse.hemisphere.parameters.txt

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/abria/TeraStitcher/issues/32, or mute the thread https://github.com/notifications/unsubscribe-auth/AIdv1q8F5euguKV_tXIpK3VjEcsHEKIbks5uD5iigaJpZM4VFw2w .

--


Giulio Iannello Preside della Facolta' Dipartimentale di Ingegneria Universita' Campus Bio-Medico di Roma v. Alvaro del Portillo, 21 00128 Roma, Italy

Tel: +39-06-22541-9602 E-mail: g.iannello@unicampus.it Fax: +39-06-22541-9609 URL: https://scholar.google.it/citations?user=L-UJxIgAAAAJ


Lalga commented 6 years ago

I apologize for my late reply! I'll answer each of your questions separately.

Thank you for your help Giulio

iannellog commented 6 years ago

From what you write, it seems that you have used the xml file provided by LAVision to "import" your dataset and then you have DIRECTLY stitched the image (using the "merge" tab of the GUI) without performing the "align" step (using the "align" tab of the GUI. Is this right? The align correct the errors in tile positions that can be happened at acquisition time. It especially correct the alignment errors along Z which seem the cause of the misalignments you observe. Pay attention that performing the align step in your case will take a significant time (even several hours, it dipends on how powerfiul is your computing platform) I attach a quick guide to use TeraStitcher that show the typical procedure to be followed. It does not cover all the possibilities of the tool and there are special cases in which a more sophisticated use of the tool is needed. However, I hope this can solve your problems. If the result is better, but there are still misalignements, consider that I have a recent version of the tool that uses a more effcient alignment algorithm. Its use, however is less immediate, so I prefer to see before if the standard version succeeds. Let me know if you get a satisfactory stitching of your image. Best.

-- Giulio

2018-07-09 19:43 GMT+02:00 Lalga notifications@github.com:

I apologize for my late reply! I'll answer each of your questions separately.

  • This is my first time using Terastitcher
  • GUI version
  • Yes, each tile has 2007 slices
  • Actually not by me, but by LAVision, the developers of the ultramicroscope we use at the lab. They built this script to run on the microscope computer
  • Good to know, I'll try to change that.
  • The file I sent you I believe is the xml file produced by the script by the LAVision developers and its the one I used to generate the image (and the test slice)
  • This is not the test slice, its a slice from the finished stitched image, I've attached the test slice below, which has many alignment issues as well. [image: image] https://user-images.githubusercontent.com/33734415/42466754-cfeea9d4-837d-11e8-8278-4d5996323553.png

Thank you for your help Giulio

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/abria/TeraStitcher/issues/32#issuecomment-403561413, or mute the thread https://github.com/notifications/unsubscribe-auth/AIdv1jQ7WjNVVeLpIXv8wYkl3mvstWVTks5uE5YkgaJpZM4VFw2w .

--


Giulio Iannello Preside della Facolta' Dipartimentale di Ingegneria Universita' Campus Bio-Medico di Roma v. Alvaro del Portillo, 21 00128 Roma, Italy

Tel: +39-06-22541-9602 E-mail: g.iannello@unicampus.it Fax: +39-06-22541-9609 URL: https://scholar.google.it/citations?user=L-UJxIgAAAAJ


Lalga commented 6 years ago

Thank you for all your help. And I believe you're right, I forgot about the Align step. I'll retry it again today, and let you know how it all goes.