Re: 3D projections

Posted by Knecht, David on
URL: http://imagej.273.s1.nabble.com/3D-projections-tp5023114p5023124.html

Hi Michael- You are correct. It is in creating rotation series with 3D project where one sees the problem.  It is not a huge problem, as the orthogonal view rarely has anything useful in it.  It is more that I am teaching this right now (online of course) and realized that I did not know how to explain that result to my students.  In the “real world” I would almost never show that view, and if I needed to, I would interpolate or use Orthogonal views.  The extra size is also not a big problem because the final result is not larger, just the memory it takes to get there and I have never run out of memory generating a rotation series.  So I am definitely not arguing someone should spend lots of time fixing the code.  Dave

Dr. David Knecht
Professor, Department of Molecular and Cell Biology
University of Connecticut
91 N. Eagleville Rd.
U-3125
Storrs, CT 06269-3125
860-486-2200

On Mar 31, 2020, at 2:30 PM, Michael Schmid <[hidden email]<mailto:[hidden email]>> wrote:

*Message sent from a system outside of UConn.*


Hi David,

which is the command that causes the problem?

I tried with the "Confocal Series" sample and remove the second color
channel, to get a 3D dataset (Image>Stacks>Tools>Make substack).
Then, pixel size in x&y is ~0.05 um, in z it is 0.6 um.

Image>Stacks>Orthogonal Views does it right.
Image>Stacks>Reslice interpolates unless you check 'avoid interpolation'
(then it shows it as if the voxels were cubes).
Image>Stacks>zProject does not rotate.

Among what I tried, the only one that shows thin slices is
Image>Stacks>3D project - seems this is the one you are talking about?

I had a look at it; the code (ij.plugin.projector) is not very easy to
understand. (The relevant function seems to be 'doOneProjectionY')
I fear that the author of the original Pascal version (Michael Castle,
probably > 20 years ago in the times of NIH Image, at the University of
Michigan Mental Health Research Institute) is not active in the ImageJ
community.

So I fear that the only easy solution is interpolation of the original
data (probably easiest with "reslice"), which will, of course, create a
much larger stack than what would be needed (9x the memory), then do the
"Image>Stacks>3D project".

Michael
___________________________________________________________________

On 31/03/2020 7:51 pm, Knecht, David wrote:
The problem is that confocal data is rarely cubic voxels.  They are nearly always much larger in z and x-y.  In the case of the data I was analyzing they were 0.06 x 0.06 x 0.5 µm and that is not unusual.  So you would expect to see the side on projection as a smear given you are looking through elongated voxels.  Dave

Dr. David Knecht
Professor, Department of Molecular and Cell Biology
University of Connecticut
91 N. Eagleville Rd.
U-3125
Storrs, CT 06269-3125
860-486-2200

On Mar 31, 2020, at 1:37 PM, Herbie <[hidden email]<mailto:[hidden email]><mailto:[hidden email]>> wrote:

*Message sent from a system outside of UConn.*


Good day Kenneth,

thanks for your reply!

"Now...why doesn't the same logic apply to image "planes"?"

With a stack of voxel-size 1x1x1 and when I use "Reslice..." (e.g.
Rotate 90deg), I'm quite happy with the result.

Not sure what's the original problem...

Regards

Herbie

:::::::::::::::::::::::::::::::::::::::::::
Am 31.03.20 um 18:38 schrieb Kenneth Sloan:
Understood.  Now...why doesn't the same logic apply to image "planes"?
--
Kenneth Sloan
[hidden email]<mailto:[hidden email]><mailto:[hidden email]> <mailto:[hidden email]>
Vision is the art of seeing what is invisible to others.

Begin forwarded message:

From: Herbie <[hidden email]<mailto:[hidden email]><mailto:[hidden email]> <mailto:[hidden email]>>

Subject: Re: 3D projections

Date: March 31, 2020 at 05:37:15 CDT

To: [hidden email]<mailto:[hidden email]><mailto:[hidden email]> <mailto:[hidden email]>

Reply-To: [hidden email]<mailto:[hidden email]><mailto:[hidden email]> <mailto:[hidden email]>


"ImageJ does not treat PIXELS as infinitely small in x and y [...]"

If you would, which were correct, you wouldn't see anything.
Therefore and in general, the cheapest interpolation is applied:
Little squares or rectangles of constant value(s).
Actually, such block images are incorrect as well, because the correct
interpolation is a totally different one that is much more costly and
cannot be realized by common display technology.

Regards

Herbie

:::::::::::::::::::::::::::::::::::::::::::
Am 30.03.20 um 23:49 schrieb Kenneth Sloan:
This has always confused me.  After all, ImageJ does not treat PIXELS
as infinitely small in x and y - so why should it consider  VOXELS to
be infinitely thin in z (but with finite  width and height)?
--
Kenneth Sloan
[hidden email]<mailto:[hidden email]><mailto:[hidden email]> <mailto:[hidden email]>
Vision is the art of seeing what is invisible to others.
--
ImageJ mailing list: https://nam10.safelinks.protection.outlook.com/?url=http%3A%2F%2Fimagej.nih.gov%2Fij%2Flist.html&amp;data=02%7C01%7Cdavid.knecht%40uconn.edu%7C4e174ec6ef5640fea6fe08d7d5a1b9ec%7C17f1a87e2a254eaab9df9d439034b080%7C0%7C0%7C637212762910092666&amp;sdata=0SPAlhIt9Ptz%2FEOxCuiTgwsNgGJFiffTcOTLDAPZw3A%3D&amp;reserved=0

--
ImageJ mailing list: https://nam10.safelinks.protection.outlook.com/?url=http%3A%2F%2Fimagej.nih.gov%2Fij%2Flist.html&amp;data=02%7C01%7Cdavid.knecht%40uconn.edu%7C4e174ec6ef5640fea6fe08d7d5a1b9ec%7C17f1a87e2a254eaab9df9d439034b080%7C0%7C0%7C637212762910102660&amp;sdata=gWwGw8cyDrMnl5M6CuSHWhGTVmhHCLUdSfDvCLkUx%2Bo%3D&amp;reserved=0

--
ImageJ mailing list: https://nam10.safelinks.protection.outlook.com/?url=http%3A%2F%2Fimagej.nih.gov%2Fij%2Flist.html&amp;data=02%7C01%7Cdavid.knecht%40uconn.edu%7C4e174ec6ef5640fea6fe08d7d5a1b9ec%7C17f1a87e2a254eaab9df9d439034b080%7C0%7C0%7C637212762910102660&amp;sdata=gWwGw8cyDrMnl5M6CuSHWhGTVmhHCLUdSfDvCLkUx%2Bo%3D&amp;reserved=0


--
ImageJ mailing list: https://nam10.safelinks.protection.outlook.com/?url=http%3A%2F%2Fimagej.nih.gov%2Fij%2Flist.html&amp;data=02%7C01%7Cdavid.knecht%40uconn.edu%7C4e174ec6ef5640fea6fe08d7d5a1b9ec%7C17f1a87e2a254eaab9df9d439034b080%7C0%7C0%7C637212762910102660&amp;sdata=gWwGw8cyDrMnl5M6CuSHWhGTVmhHCLUdSfDvCLkUx%2Bo%3D&amp;reserved=0


--
ImageJ mailing list: https://nam10.safelinks.protection.outlook.com/?url=http%3A%2F%2Fimagej.nih.gov%2Fij%2Flist.html&amp;data=02%7C01%7Cdavid.knecht%40uconn.edu%7C4e174ec6ef5640fea6fe08d7d5a1b9ec%7C17f1a87e2a254eaab9df9d439034b080%7C0%7C0%7C637212762910102660&amp;sdata=gWwGw8cyDrMnl5M6CuSHWhGTVmhHCLUdSfDvCLkUx%2Bo%3D&amp;reserved=0


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html