Import EDAX - transformations again

48 views
Skip to first unread message

Michał Kalkowski

unread,
Dec 11, 2019, 4:18:29 PM12/11/19
to dream3d-users
Hello, 

I am aware of the discussions regarding the choice for transformations that bring the crystal and the sample systems together, especially in (https://groups.google.com/d/msg/dream3d-users/r3r0H2HUs1g/B16NgleuDAAJ). But the issue is still partly solved for me.

I took the Import Edax data prebuilt pipeline and requested a DREAM3D file on top of the png with IPF color (Image Writer filter). Please see the resulting images:
png file:

delete_me.png

screenshot from DREAM3d file (Paraview displaying IPF color):

Selection_146.png


The two are clearly flipped vertically. In the pipeline (attached), I disabled both the sample and the Euler reference frame rotations intentionally. I just wanted to make sure that I get the same image, as a sanity check. As you may see above this has failed.
Possible reasons - does PNG have the origin at the top left corner? I could not find any evidence of this.

I would also like to ask for a final stance on the default Euler transformation for EDAX mentioned in the thread above. It seemed that the needed rotation is 90 clockwise, but with respect to the right-hand rule this would be -90. Is the implementation in DREAM3D correct? The descriptions of transformations are still sort of ambiguous since the right-hand rule is assumed most often.

Finally, in the thread linked above it is mentioned that the image is by default rotated by 180 degrees with respect to the sample. My understanding is that if we would like to reconstruct the microstructure from EBSD and compare it to the sample, after the recommended transformations we need to rotate both the crystal and the sample reference frames by 180 degrees, see the image below:

Selection_147.png


I would be grateful if anyone could clarify these issues.


Many thanks,


Michal

---

Dr Michał K. Kalkowski

Research Associate

Non-Destructive Evaluation Group

Imperial College London

SW7 2AZ

United Kingdom

---



read_EDAX.json

Michael Jackson

unread,
Dec 12, 2019, 10:44:00 AM12/12/19
to Michał Kalkowski, dream3d-users

Dear Dr. Kalkowski,

    Your screen shot from ParaView has the (0,0) origin in the lower left. As software engineers we like to have graphic and image (0,0) origins in the upper left because for us (English is our primary language) it is easier to mentally think about memory layout from left to right and from top to bottom. PNG will just display with an origin in the upper left. Some TIFF writers will allow you to set the origin into any corner. We try to be consistent in our image writing so that you, as the user, can look at an image and look at ParaView and visually “see” the same image. Unfortunately, ParaView’s default views like to put the (0,0) origin in the lower left (which makes sense from a scientific data point of view) so a basic vertical flip of the 3D object is usually required.

 

--

Mike Jackson

 

From: <dream3...@googlegroups.com> on behalf of Michał Kalkowski <kalko...@gmail.com>
Date: Wednesday, December 11, 2019 at 4:18 PM
To: dream3d-users <dream3...@googlegroups.com>
Subject: Import EDAX - transformations again

 

Hello, 

 

I am aware of the discussions regarding the choice for transformations that bring the crystal and the sample systems together, especially in (https://groups.google.com/d/msg/dream3d-users/r3r0H2HUs1g/B16NgleuDAAJ). But the issue is still partly solved for me.

 

I took the Import Edax data prebuilt pipeline and requested a DREAM3D file on top of the png with IPF color (Image Writer filter). Please see the resulting images:

png file:

screenshot from DREAM3d file (Paraview displaying IPF color):

 

The two are clearly flipped vertically. In the pipeline (attached), I disabled both the sample and the Euler reference frame rotations intentionally. I just wanted to make sure that I get the same image, as a sanity check. As you may see above this has failed.

Possible reasons - does PNG have the origin at the top left corner? I could not find any evidence of this.

 

I would also like to ask for a final stance on the default Euler transformation for EDAX mentioned in the thread above. It seemed that the needed rotation is 90 clockwise, but with respect to the right-hand rule this would be -90. Is the implementation in DREAM3D correct? The descriptions of transformations are still sort of ambiguous since the right-hand rule is assumed most often.

 

Finally, in the thread linked above it is mentioned that the image is by default rotated by 180 degrees with respect to the sample. My understanding is that if we would like to reconstruct the microstructure from EBSD and compare it to the sample, after the recommended transformations we need to rotate both the crystal and the sample reference frames by 180 degrees, see the image below:

 

I would be grateful if anyone could clarify these issues.

 

Many thanks,

 

Michal

---

Dr Michał K. Kalkowski

Research Associate

Non-Destructive Evaluation Group

Imperial College London

SW7 2AZ

United Kingdom

---

 

 

--
You received this message because you are subscribed to the Google Groups "dream3d-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dream3d-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dream3d-users/a91813f5-7af5-4ee8-8955-717e3a2622d2%40googlegroups.com.

Michał Kalkowski

unread,
Dec 12, 2019, 11:56:19 AM12/12/19
to dream3d-users
Many thanks, Mike - that is very helpful. My suspicion was correct, then. I do not mind the origin being in the top or the bottom corner, just need to have it well-defined. Not sure about the importance of the primary language, though...;) 

Should I assume that the answers to my two further questions are affirmative? Let me recall these points below:
1) The recommended transformation for EDAX data is -90 degrees around <001> according to the right-hand rule? (not +90, as mentioned in the software)
2) The image is rotated by 180 degrees with respect to the physical specimen, and this has nothing to do with the vertical flip between PNG file and the Paraview display mentioned above (as mentioned in the lecture notes by A.D. Rollett, screenshot of which I included before). Dream3D does not rotate the image 'behind the scenes' to match the physical markings on the specimen. If one wishes to align the reconstructed microstructure with the physical markings, they would need to apply 180 degrees rotation to both the crystal and sample reference frames using appropriate filters in Dream3D. Is that correct?

I understand that these questions may seem related to the merits of the EBSD measurements, but my actual intention is to make sure that I know what Dream3D does when processing the data so that I can be certain about the results of the reconstruction, reference frames and the origins.

Many thanks again!
Michal

To unsubscribe from this group and stop receiving emails from it, send an email to dream3...@googlegroups.com.

Michael Jackson

unread,
Dec 13, 2019, 3:17:32 PM12/13/19
to Michał Kalkowski, dream3d-users

Dear Dr. Kalkowski,

   I am awaiting some input from Dr. Groeber for the other questions that you have asked. I don’t want to hand out any false information and then really mess something up.

 

--

Michael Jackson | Owner, President

      BlueQuartz Software

[e] mike.j...@bluequartz.net

[w] www.bluequartz.net

 

From: <dream3...@googlegroups.com> on behalf of Michał Kalkowski <kalko...@gmail.com>
Date: Thursday, December 12, 2019 at 11:56 AM
To: dream3d-users <dream3...@googlegroups.com>
Subject: Re: Import EDAX - transformations again

 

Many thanks, Mike - that is very helpful. My suspicion was correct, then. I do not mind the origin being in the top or the bottom corner, just need to have it well-defined. Not sure about the importance of the primary language, though...;) 

 

Should I assume that the answers to my two further questions are affirmative? Let me recall these points below:

1) The recommended transformation for EDAX data is -90 degrees around <001> according to the right-hand rule? (not +90, as mentioned in the software)

2) The image is rotated by 180 degrees with respect to the physical specimen, and this has nothing to do with the vertical flip between PNG file and the Paraview display mentioned above (as mentioned in the lecture notes by A.D. Rollett, screenshot of which I included before). Dream3D does not rotate the image 'behind the scenes' to match the physical markings on the specimen. If one wishes to align the reconstructed microstructure with the physical markings, they would need to apply 180 degrees rotation to both the crystal and sample reference frames using appropriate filters in Dream3D. Is that correct?

 

I understand that these questions may seem related to the merits of the EBSD measurements, but my actual intention is to make sure that I know what Dream3D does when processing the data so that I can be certain about the results of the reconstruction, reference frames and the origins.

 

Many thanks again!

Michal

 


On Thursday, December 12, 2019 at 3:44:00 PM UTC, Michael Jackson wrote:

Dear Dr. Kalkowski,

    Your screen shot from ParaView has the (0,0) origin in the lower left. As software engineers we like to have graphic and image (0,0) origins in the upper left because for us (English is our primary language) it is easier to mentally think about memory layout from left to right and from top to bottom. PNG will just display with an origin in the upper left. Some TIFF writers will allow you to set the origin into any corner. We try to be consistent in our image writing so that you, as the user, can look at an image and look at ParaView and visually “see” the same image. Unfortunately, ParaView’s default views like to put the (0,0) origin in the lower left (which makes sense from a scientific data point of view) so a basic vertical flip of the 3D object is usually required.

 

--

Mike Jackson

 

From: <dream3...@googlegroups.com> on behalf of Michał Kalkowski <kalko...@gmail.com>
Date: Wednesday, December 11, 2019 at 4:18 PM
To: dream3d-users <dream3...@googlegroups.com>
Subject: Import EDAX - transformations again

 

Hello, 

 

I am aware of the discussions regarding the choice for transformations that bring the crystal and the sample systems together, especially in (https://groups.google.com/d/msg/dream3d-users/r3r0H2HUs1g/B16NgleuDAAJ). But the issue is still partly solved for me.

 

I took the Import Edax data prebuilt pipeline and requested a DREAM3D file on top of the png with IPF color (Image Writer filter). Please see the resulting images:

png file:

Image removed by sender.

screenshot from DREAM3d file (Paraview displaying IPF color):

Image removed by sender.

 

The two are clearly flipped vertically. In the pipeline (attached), I disabled both the sample and the Euler reference frame rotations intentionally. I just wanted to make sure that I get the same image, as a sanity check. As you may see above this has failed.

Possible reasons - does PNG have the origin at the top left corner? I could not find any evidence of this.

 

I would also like to ask for a final stance on the default Euler transformation for EDAX mentioned in the thread above. It seemed that the needed rotation is 90 clockwise, but with respect to the right-hand rule this would be -90. Is the implementation in DREAM3D correct? The descriptions of transformations are still sort of ambiguous since the right-hand rule is assumed most often.

 

Finally, in the thread linked above it is mentioned that the image is by default rotated by 180 degrees with respect to the sample. My understanding is that if we would like to reconstruct the microstructure from EBSD and compare it to the sample, after the recommended transformations we need to rotate both the crystal and the sample reference frames by 180 degrees, see the image below:

Image removed by sender.

 

I would be grateful if anyone could clarify these issues.

 

Many thanks,

 

Michal

---

Dr Michał K. Kalkowski

Research Associate

Non-Destructive Evaluation Group

Imperial College London

SW7 2AZ

United Kingdom

---

 

 

--
You received this message because you are subscribed to the Google Groups "dream3d-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dream3...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dream3d-users/a91813f5-7af5-4ee8-8955-717e3a2622d2%40googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "dream3d-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to dream3d-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dream3d-users/036ccc08-08f4-4be8-b4f7-5f484b2846e2%40googlegroups.com.

Reply all
Reply to author
Forward
0 new messages