Stitching output has missing pixels

[MChapman]

Hello,

I have been fighting Stephan's stitching plugin for a few months now and I just wanted to let you guys know what my issues have been if there is a hope at solving them.  I really like the stitching algorithm, there are just two errors that I have been experiencing.

1) When stitching more than about 50-100 images, the first or first and second tile don't find an adequate alignment and are left out of the stitch and assigned a position of (0,0) in the  TileConfiguration.txt file.

To correct this, I just run a second stitch of the first 2-3 rows of the montage.  Then I use the position of an image in the middle of the bottom row of this stitch as a reference and rewrite the TileConfiguration.txt file positions for the full stitch.  I can then rerun the stitching program without "compute overlap" with the repaired TileConfiguration.txt and this fixes the first error.  A bit of a nuisance, but it works.

2) When I rerun the stitches with the repaired TileConfiguration.txt, some of the pixels (sometimes a single pixel, sometimes a block of 20x300) come back as 0 in the stitch.  There are no gaps in the overlap of the images from the TileConfiguration.txt file, and if I rerun the stitch again the 0 pixels show up in different locations.

My current fix is to just run multiple restitches and then take a max of a stack of the stitches and do that until there are no 0 pixels.  Again this is a nuisance but I can get my data stitched with it.

Is there anything that I am doing wrong or is anybody else having these issues?

Mike

edit* PS...If example images would be helpful, just let me know and I can post them or post a link to them.
I reposted this correctly this time to send it out to the mailing list
MIke

[Majurski, Michael Paul (Fed)]

Mike,

I have run into that missing output tile bug myself a few times and so far I have not seen neither a fix nor speculation as to the cause.

At times I have had the first few tiles end up missing from the output image when I know I am using the plugin correctly. So your bug 1 is not a user error. However, I have never has small blocks of the output image go missing (bug 2 in your email) that were not in the upper left corner (tile 1, tile 1-2). So I cannot comment on that apparent bug.

Another resource (if you have not found it already) would be the Fiji forum: http://forum.imagej.net/,  where there have been a few discussion related to Stephan's stitching plugin.

However, there are other stitching programs to try which might result in less effort for you. The one I was involved in creating (MIST https://pages.nist.gov/MIST/ ) requires a regular grid of images (basically roughly constant overlap), but it might be worth giving it a try.

Project Page: https://pages.nist.gov/MIST/ 
Source: https://github.com/usnistgov/MIST
Wiki: https://github.com/usnistgov/MIST/wiki
Install Guide: https://github.com/usnistgov/MIST/wiki/Install-Guide

Best of luck,
Michael Majurski


-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of MChapman
Sent: Wednesday, December 07, 2016 17:14
To: [hidden email]
Subject: Stitching output has missing pixels

Hello,

I have been fighting Stephan's stitching plugin for a few months now and I just wanted to let you guys know what my issues have been if there is a hope at solving them.  I really like the stitching algorithm, there are just two errors that I have been experiencing.

1) When stitching more than about 50-100 images, the first or first and second tile don't find an adequate alignment and are left out of the stitch and assigned a position of (0,0) in the  TileConfiguration.txt file.

To correct this, I just run a second stitch of the first 2-3 rows of the montage.  Then I use the position of an image in the middle of the bottom row of this stitch as a reference and rewrite the TileConfiguration.txt file positions for the full stitch.  I can then rerun the stitching program without "compute overlap" with the repaired TileConfiguration.txt and this fixes the first error.  A bit of a nuisance, but it works.

2) When I rerun the stitches with the repaired TileConfiguration.txt, some of the pixels (sometimes a single pixel, sometimes a block of 20x300) come back as 0 in the stitch.  There are no gaps in the overlap of the images from the TileConfiguration.txt file, and if I rerun the stitch again the 0 pixels show up in different locations.

My current fix is to just run multiple restitches and then take a max of a stack of the stitches and do that until there are no 0 pixels.  Again this is a nuisance but I can get my data stitched with it.

Is there anything that I am doing wrong or is anybody else having these issues?

Mike

edit* PS...If example images would be helpful, just let me know and I can post them or post a link to them.
I reposted this correctly this time to send it out to the mailing list MIke



--
View this message in context: http://imagej.1557.x6.nabble.com/Stitching-output-has-missing-pixels-tp5017714.html
Sent from the ImageJ mailing list archive at Nabble.com.

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

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

[Aryeh Weiss]

I posted on this issue , most recently:
https://list.nih.gov/cgi-bin/wa.exe?A2=ind1610&L=IMAGEJ&F=&S=&P=41767

I characterized what happens as the stitching is done over larger areas.

By looking at the tiles near the blank area, you can get a decent
estimate of where to put the missing tiles. However, it would be nice to
not have to do that.

Best regards
--aryeh

On 08/12/2016 16:59, Majurski, Michael Paul (Fed) wrote:

> Mike,
>
> I have run into that missing output tile bug myself a few times and so far I have not seen neither a fix nor speculation as to the cause.
>
> At times I have had the first few tiles end up missing from the output image when I know I am using the plugin correctly. So your bug 1 is not a user error. However, I have never has small blocks of the output image go missing (bug 2 in your email) that were not in the upper left corner (tile 1, tile 1-2). So I cannot comment on that apparent bug.
>
> Another resource (if you have not found it already) would be the Fiji forum: http://forum.imagej.net/,  where there have been a few discussion related to Stephan's stitching plugin.
>
> However, there are other stitching programs to try which might result in less effort for you. The one I was involved in creating (MIST https://pages.nist.gov/MIST/ ) requires a regular grid of images (basically roughly constant overlap), but it might be worth giving it a try.
>
> Project Page: https://pages.nist.gov/MIST/
> Source: https://github.com/usnistgov/MIST
> Wiki: https://github.com/usnistgov/MIST/wiki
> Install Guide: https://github.com/usnistgov/MIST/wiki/Install-Guide
>
> Best of luck,
> Michael Majurski
>
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of MChapman
> Sent: Wednesday, December 07, 2016 17:14
> To: [hidden email]
> Subject: Stitching output has missing pixels
>
> Hello,
>
> I have been fighting Stephan's stitching plugin for a few months now and I just wanted to let you guys know what my issues have been if there is a hope at solving them.  I really like the stitching algorithm, there are just two errors that I have been experiencing.
>
> 1) When stitching more than about 50-100 images, the first or first and second tile don't find an adequate alignment and are left out of the stitch and assigned a position of (0,0) in the  TileConfiguration.txt file.
>
> To correct this, I just run a second stitch of the first 2-3 rows of the montage.  Then I use the position of an image in the middle of the bottom row of this stitch as a reference and rewrite the TileConfiguration.txt file positions for the full stitch.  I can then rerun the stitching program without "compute overlap" with the repaired TileConfiguration.txt and this fixes the first error.  A bit of a nuisance, but it works.
>
> 2) When I rerun the stitches with the repaired TileConfiguration.txt, some of the pixels (sometimes a single pixel, sometimes a block of 20x300) come back as 0 in the stitch.  There are no gaps in the overlap of the images from the TileConfiguration.txt file, and if I rerun the stitch again the 0 pixels show up in different locations.
>
> My current fix is to just run multiple restitches and then take a max of a stack of the stitches and do that until there are no 0 pixels.  Again this is a nuisance but I can get my data stitched with it.
>
> Is there anything that I am doing wrong or is anybody else having these issues?
>
> Mike
>
> edit* PS...If example images would be helpful, just let me know and I can post them or post a link to them.
> I reposted this correctly this time to send it out to the mailing list MIke
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Stitching-output-has-missing-pixels-tp5017714.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Aryeh Weiss
Faculty of Engineering
Bar Ilan University
Ramat Gan 52900 Israel

Ph:  972-3-5317638
FAX: [phone]

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

[MChapman]

Aryeh,

It is interesting that the error only shows up when there are more than 90 tiles because I have seen the error flag in the Log as it being a poor alignment of the first two tiles.

I also do not understand why it would throw out the first tile if it is a bad match...shouldn't the first tile be the reference and if it was a bad match between the first two that it would disregard the second tile?

I think the bug lies in that part of the program.

Mike

[Aryeh Weiss]

I think that there is always a problem with placement of the first tile,
but only  when the error grows large enough does it not get place in the
fusion. This is only a guess.

This is the reason I did an entire series of sizes -- to see it develop.

Best regards
--aryeh

On 08/12/2016 21:56, MChapman wrote:

> Aryeh,
>
> It is interesting that the error only shows up when there are more than 90
> tiles because I have seen the error flag in the Log as it being a poor
> alignment of the first two tiles.
>
> I also do not understand why it would throw out the first tile if it is a
> bad match...shouldn't the first tile be the reference and if it was a bad
> match between the first two that it would disregard the second tile?
>
> I think the bug lies in that part of the program.
>
> Mike
>
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Stitching-output-has-missing-pixels-tp5017714p5017725.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Aryeh Weiss
Faculty of Engineering
Bar Ilan University
Ramat Gan 52900 Israel

Ph:  972-3-5317638
FAX: [phone]

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

[MChapman]

I would have agreed until yesterday, when I had the algorithm misplace the first tile in a 1x2 tile stitch.

The fusion just displayed the second tile.  That is when I actually saw it kick it out in the Log.

Mike

[ctrueden]

Hi Mike,

Sounds like you found a nice, small, reproducible test case.  Perhaps you
could share the data + macro illustrating the issue? It could help the
plugin maintainers to investigate and fix thr issue more effectively.

Regards,
Curtis


On Dec 8, 2016 3:08 PM, "MChapman" <[hidden email]> wrote:

I would have agreed until yesterday, when I had the algorithm misplace the
first tile in a 1x2 tile stitch.

The fusion just displayed the second tile.  That is when I actually saw it
kick it out in the Log.

Mike



--
View this message in context: http://imagej.1557.x6.nabble.
com/Stitching-output-has-missing-pixels-tp5017714p5017728.html
Sent from the ImageJ mailing list archive at Nabble.com.

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

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

[MChapman]

Curtis,

I would be happy to share, and I have been trying to figure out how to upload these two images (16bit, 1292x968 .tif) but whenever I try to shrink the images or change the file type this stitching error doesn't occur anymore.  I also need to get these images cleared for release which may take a few weeks to a month.

If there is somewhere that I could post the pictures in their raw form I will.

The macro for the stitch is:
        run("Grid/Collection stitching", "type=[Grid: snake by rows] order=[Right & Down                ] grid_size_x=2 grid_size_y=1 tile_overlap=20 first_file_index_i=0 directory=[E:\ImageJ mailing list] file_names=1000{i}.tif output_textfile_name=TileConfiguration.txt fusion_method=[Linear Blending] regression_threshold=0.30 max/avg_displacement_threshold=2.50 absolute_displacement_threshold=3.50 compute_overlap subpixel_accuracy computation_parameters=[Save computation time (but use more RAM)] image_output=[Fuse and display]");

and the resulting log file output is:
Stitching internal version: 1.2
Loading (0, 0): E:\\ImageJ mailing list\10000.tif ...
1292x968px, channels=1, timepoints=1 (80 ms)
Loading (1, 0): E:\\ImageJ mailing list\10001.tif ...
1292x968px, channels=1, timepoints=1 (70 ms)
10000.tif[1] <- 10001.tif[1]: (1037.8536, -1.7621746) correlation (R)=0.29953933 (150 ms)
ERROR: Error: No correlated tiles found, setting the first tile to (0, 0).
 number of tiles = 1
Finished registration process (150 ms).
10000.tif: [3,3](AffineTransform[[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]]) 1.7976931348623157E308
Writing registered TileConfiguration: E:ImageJ mailing list\TileConfiguration.registered.txt
Fuse & Display ...
Finished fusion (70 ms)
Finished ... (370 ms)

Hi, I am happy to have a look at it, I am sure this is a quick fix. You could upload it to google drive as a zip and share the document, this does work with pretty big files.

Sorry for the delay, I was on holidays.

All the best,
Stephan
---

Dr. Stephan Preibisch
Group Leader

Berlin Institute of Medical Systems Biology of the MDC
Building 89, 1.08b

email: [hidden email]<mailto:[hidden email]>
web: http://www.preibisch.net/<http://fly.mpi-cbg.de/preibisch>

On Dec 9, 2016, at 1:47 PM, MChapman <[hidden email]<mailto:[hidden email]>> wrote:

Curtis,

I would be happy to share, and I have been trying to figure out how to
upload these two images (16bit, 1292x968 .tif) but whenever I try to shrink
the images or change the file type this stitching error doesn't occur
anymore.

If there is somewhere that I could post the pictures in their raw form I
will.

The macro for the stitch is:
run("Grid/Collection stitching", "type=[Grid: snake by rows] order=[Right &
Down                ] grid_size_x=2 grid_size_y=1 tile_overlap=20
first_file_index_i=0 directory=[E:\ImageJ mailing list]
file_names=1000{i}.tif output_textfile_name=TileConfiguration.txt
fusion_method=[Linear Blending] regression_threshold=0.30
max/avg_displacement_threshold=2.50 absolute_displacement_threshold=3.50
compute_overlap subpixel_accuracy computation_parameters=[Save computation
time (but use more RAM)] image_output=[Fuse and display]");

and the resulting log file output is:
Stitching internal version: 1.2
Loading (0, 0): E:\\ImageJ<smb://ImageJ> mailing list\10000.tif ...
1292x968px, channels=1, timepoints=1 (80 ms)
Loading (1, 0): E:\\ImageJ<smb://ImageJ> mailing list\10001.tif ...
1292x968px, channels=1, timepoints=1 (70 ms)
10000.tif[1] <- 10001.tif[1]: (1037.8536, -1.7621746) correlation
(R)=0.29953933 (150 ms)
ERROR: Error: No correlated tiles found, setting the first tile to (0, 0).
number of tiles = 1
Finished registration process (150 ms).
10000.tif: [3,3](AffineTransform[[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]])
1.7976931348623157E308
Writing registered TileConfiguration: E:ImageJ mailing
list\TileConfiguration.registered.txt
Fuse & Display ...
Finished fusion (70 ms)
Finished ... (370 ms)




--
View this message in context: http://imagej.1557.x6.nabble.com/Stitching-output-has-missing-pixels-tp5017714p5017737.html
Sent from the ImageJ mailing list archive at Nabble.com<http://nabble.com>.

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


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