A wrong result from terastitcher

[fangchunyu]

I want to test a group of data which includes 1 column and 2 rows. Each of stack has 20482048391 voxels.

With hierarchy like this, I can import these images correctly. But the merged results are still 20592048360. The overlap seems to be 100% even when I set a 205 overlap voxels. It should be something like this:

But the terastitcher merged this: I wonder why they overlap together and how can I avoid this by setting parameters.

[pawlowska]

what does your test slice look like?

[iannellog]

In order to help you I have to better understand your dataset and which commands are you issueing.

From your mail I understand that:

• you have a 2 rows x 1 column tile matrix
• each tile is 2048 x 2048 x 391
• color depth is 16 bits

I would need to know:

• which is the voxel size in um
• which is the nominal overlap between tiles
• which is the exact commands you are issueing or, alternatively, if you are using the GUI, which are the parameters you have used I need also to receive all the xml files that TeraStitcher has generated.

As a preliminary observation, your images seem to have poor contrast. The slices you show have very different Z values and therefore it is difficult to draw conclusions, but if there are no clearly visible common structures between adjacent tiles it is difficult to correct alignment errors and the alignment algorithm may provide wrong results.

Best.

-- Giulio

Il giorno ven 14 dic 2018 alle ore 04:23 fangchunyu < notifications@github.com> ha scritto:

I want to test a group of data which includes 1 column and 2 rows. Each of stack has 20482048391 voxels. [image: image] https://user-images.githubusercontent.com/16661193/49980875-49015e00-ff90-11e8-89af-8003b9f6f527.png

[image: image] https://user-images.githubusercontent.com/16661193/49980887-54548980-ff90-11e8-821c-abff6cc284e1.png [image: image] https://user-images.githubusercontent.com/16661193/49980897-60404b80-ff90-11e8-9744-559900531fbb.png With hierarchy like this, I can import these images correctly. But the merged results are still 20592048360. The overlap seems to be 100% even when I set a 205 overlap voxels. It should be something like this: [image: image] https://user-images.githubusercontent.com/16661193/49981012-f96f6200-ff90-11e8-980e-b462ab988ee2.png

But the terastitcher merged this: [image: image] https://user-images.githubusercontent.com/16661193/49981361-836bfa80-ff92-11e8-84ce-837b8ea3daa0.png I wonder why they overlap together and how can I avoid this by setting parameters.

— 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/41, or mute the thread https://github.com/notifications/unsubscribe-auth/AIdv1p09174fIjJ5yvoUU2Fz0vT0HKT0ks5u4xmUgaJpZM4ZS0eG .

--

---

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

---

[fangchunyu]

what does your test slice look like?

Well I tested a volume of mouse kidney, just like half of a ball

[fangchunyu]

In order to help you I have to better understand your dataset and which commands are you issueing. From your mail I understand that: - you have a 2 rows x 1 column tile matrix - each tile is 2048 x 2048 x 391 - color depth is 16 bits I would need to know: - which is the voxel size in um - which is the nominal overlap between tiles - which is the exact commands you are issueing or, alternatively, if you are using the GUI, which are the parameters you have used I need also to receive all the xml files that TeraStitcher has generated. As a preliminary observation, your images seem to have poor contrast. The slices you show have very different Z values and therefore it is difficult to draw conclusions, but if there are no clearly visible common structures between adjacent tiles it is difficult to correct alignment errors and the alignment algorithm may provide wrong results. Best. … -- Giulio Il giorno ven 14 dic 2018 alle ore 04:23 fangchunyu < notifications@github.com> ha scritto: I want to test a group of data which includes 1 column and 2 rows. Each of stack has 20482048391 voxels. [image: image] https://user-images.githubusercontent.com/16661193/49980875-49015e00-ff90-11e8-89af-8003b9f6f527.png [image: image] https://user-images.githubusercontent.com/16661193/49980887-54548980-ff90-11e8-821c-abff6cc284e1.png [image: image] https://user-images.githubusercontent.com/16661193/49980897-60404b80-ff90-11e8-9744-559900531fbb.png With hierarchy like this, I can import these images correctly. But the merged results are still 20592048360. The overlap seems to be 100% even when I set a 205 overlap voxels. It should be something like this: [image: image] https://user-images.githubusercontent.com/16661193/49981012-f96f6200-ff90-11e8-980e-b462ab988ee2.png But the terastitcher merged this: [image: image] https://user-images.githubusercontent.com/16661193/49981361-836bfa80-ff92-11e8-84ce-837b8ea3daa0.png I wonder why they overlap together and how can I avoid this by setting parameters. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#41>, or mute the thread https://github.com/notifications/unsubscribe-auth/AIdv1p09174fIjJ5yvoUU2Fz0vT0HKT0ks5u4xmUgaJpZM4ZS0eG .

____ 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

---

The voxel size is 1.51.51.5 in three dimensions and the overlap is 10% of image, which means almost 205 pixels. I used the GUI and it generated an xml file. I have tried to read this file and I find the default overlap is 50%. Then I modified the displacement to 1843 and the result become much better. Yeah my images have very different Z values and no clearly visible common structures between adjacent tiles, but for example, if I have a 5 rows* 6 columns tile matrix and just part of them have common structures, will it output a result that have different correctness in different area?