Grid/Collection Stitch
Locked
 
|
Hi All,
I'm new to FIJI, and I was troubleshooting the Grid stitching plugin. It was able to produce a stitched image, but some areas are aligned incorrectly. For example, the nuclei from overlapping constituent images are juxtaposed in the fused image instead of merged together (upper right part of the neurosphere, adjacent to the 2 bubbles). The images are in the following folder: https://www.dropbox.com/sh/8idzyd0rh3qt3in/AADUA93Kd-31bK-pzzwCgt50a?dl=0 "C3-08C6 _A_comp 10x Fused" is the stitched image "08C6 _A_Hoechst_10x_13" is the corresponding original I've tried to increase the regression threshold and decrease the displacement threshold, but without any luck. Is there anything I can change in the way I take the pictures or stitch them together? Is there such a thing as too much overlap in the individual pictures? Thank you for your help. Sincerely, Cheuk -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Re: Grid/Collection Stitch
|
|
Cheuk,
In the past when I have had trouble getting the grid/collection stitching plugin to assemble a satisfactory image I have set the regression threshold very low (0.1). I have found that if you provide a correct image overlap (that is also uniform between all images) the plugin will sometimes generate a correct image. ~Michael Majurski -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Cheuk Tam Sent: Friday, October 16, 2015 1:53 PM To: [hidden email] Subject: Grid/Collection Stitch Hi All, I'm new to FIJI, and I was troubleshooting the Grid stitching plugin. It was able to produce a stitched image, but some areas are aligned incorrectly. For example, the nuclei from overlapping constituent images are juxtaposed in the fused image instead of merged together (upper right part of the neurosphere, adjacent to the 2 bubbles). The images are in the following folder: https://www.dropbox.com/sh/8idzyd0rh3qt3in/AADUA93Kd-31bK-pzzwCgt50a?dl=0 "C3-08C6 _A_comp 10x Fused" is the stitched image "08C6 _A_Hoechst_10x_13" is the corresponding original I've tried to increase the regression threshold and decrease the displacement threshold, but without any luck. Is there anything I can change in the way I take the pictures or stitch them together? Is there such a thing as too much overlap in the individual pictures? Thank you for your help. Sincerely, Cheuk -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
|
|
In reply to this post by Cheuk Tam-2
Dear Dr. Majurski,
Thank you for your suggestions. Lowering the regression threshold gave better alignment to parts of the image. Unfortunately, the nuclei were still misaligned. I was able to have proper alignment when stitching the channels (red for Beta-3 Tubulin and blue for Hoechst) separately. However, it was difficult merge the two channels together after stitching because the fused images are of different sizes. I was wondering there is an issue with registering different channels together. If so, will converting the channels to timepoints (perhaps by designating channels as tandem timepoints) help? Is there a way to do this in MIST? Thank you for your help. Sincerely, Cheuk -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Re: Grid/Collection Stitch
|
|
Cheuk,
If you stitch multiple channels such that each channel gets registered independently then it will almost always result in stitched images of different sizes. The method I usually use is to select a registration channel, typically the one with the most information. I like to use phase contrast, but whatever channel has the more foreground information in the overlap regions between image should be used. By selecting the channel with the most foreground information in the overlapping regions you give the stitching algorithm the best possible chance of stitching correctly. I then take the registration information from the registration channel and assemble the other channels using that information. That results in all of the stitched images being the same size and exactly registered with each other. I do not know of a way to do this using the Grid/Collection stitching tool. I usually let the Grid/Collection tool generate the global image positions ("TileConfiguration.registered.txt") and then I parse and assemble that file using a function I have written in Matlab. This allows me to stitch a registration channel, have the global image positions file generated, and then parse and assemble all of the channels I am interested in using the global image positions. If you are inclined to try the MIST stitching tool there is a method to assemble from metadata which allows you to stitch a registration channel and then use that to assemble secondary channels. This link explains how: https://github.com/NIST-ISG/MIST/wiki/FAQ#stitching-multiple-channels . I am not sure that I understand what you meant by convert the channels to timepoints. From what I have read (from http://fiji.sc/Image_Stitching#Register_different_channels_to_each_other ) it seems that only the pairwise stitching from the Grid/Collection tool will handle time series. If you can post the images to be stitched on dropbox (with the previous results) I would be happy to play with them and figure out what I can get MIST to produce and give you a script. Thanks, Michael Majurski -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Cheuk Tam Sent: Wednesday, October 21, 2015 2:47 PM To: [hidden email] Subject: Re: Grid/Collection Stitch Dear Dr. Majurski, Thank you for your suggestions. Lowering the regression threshold gave better alignment to parts of the image. Unfortunately, the nuclei were still misaligned. I was able to have proper alignment when stitching the channels (red for Beta-3 Tubulin and blue for Hoechst) separately. However, it was difficult merge the two channels together after stitching because the fused images are of different sizes. I was wondering there is an issue with registering different channels together. If so, will converting the channels to timepoints (perhaps by designating channels as tandem timepoints) help? Is there a way to do this in MIST? Thank you for your help. Sincerely, Cheuk -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
|
|
In reply to this post by Cheuk Tam-2
Dear Dr. Majurski,
Thank you for your response and generous offer to help. It would be amazing to use MIST to construct a fused image from metadata from Grid/Collection stitching of the optimal (Tuj1) channel. I've uploaded constituent images for a neurosphere to the following dropbox folder: https://www.dropbox.com/sh/8idzyd0rh3qt3in/AADUA93Kd-31bK-pzzwCgt50a?dl=0 The 08C6 Tuj1 and 08C6 Tuj1_Hoechst Overlay contain images taken in sequential, horizontal order starting at the upper left corner of the field of view. There are 5 rows of 4. The "Previous Results" folder contain an image from stitching composite Tuj1/Hoechst pictures ("08C6_A_comp 10x fused") and from Tuj1 images alone ("08C6_A_Tuj1 10x fused"). The subfolder "Blurry Nuclei" contains an image of the Hoechst channel of the fused composite showing the blurry nuclei (as described last time) and an image of the corresponding constituent image. Thank you again so much for your help. Sincerely, Cheuk -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
Re: Grid/Collection Stitch
|
|
It appears that your images do not have a uniform overlap between them. Some of them have approximately 25% overlap and some closer to 50%.
Because of that MIST cannot stitch these images. One of the core MIST assumptions is a uniform grid. I have been unable to produce a viable stitched image using the Grid/Collection stitching plugin which mirrors your experience. The last suggestions I have would be to use TrakEm2 to manually or semi-manually register the images into a composite. Main site: http://fiji.sc/TrakEM2 Tutorials: http://fiji.sc/TrakEM2_tutorials A User manual: https://www.ini.uzh.ch/~acardona/trakem2_manual.html Best of luck Michael Majurski -----Original Message----- From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Cheuk Tam Sent: Thursday, October 22, 2015 7:53 PM To: [hidden email] Subject: Re: Grid/Collection Stitch Dear Dr. Majurski, Thank you for your response and generous offer to help. It would be amazing to use MIST to construct a fused image from metadata from Grid/Collection stitching of the optimal (Tuj1) channel. I've uploaded constituent images for a neurosphere to the following dropbox folder: https://www.dropbox.com/sh/8idzyd0rh3qt3in/AADUA93Kd-31bK-pzzwCgt50a?dl=0 The 08C6 Tuj1 and 08C6 Tuj1_Hoechst Overlay contain images taken in sequential, horizontal order starting at the upper left corner of the field of view. There are 5 rows of 4. The "Previous Results" folder contain an image from stitching composite Tuj1/Hoechst pictures ("08C6_A_comp 10x fused") and from Tuj1 images alone ("08C6_A_Tuj1 10x fused"). The subfolder "Blurry Nuclei" contains an image of the Hoechst channel of the fused composite showing the blurry nuclei (as described last time) and an image of the corresponding constituent image. Thank you again so much for your help. Sincerely, Cheuk -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html |
«
Return to ImageJ
|
1 view|%1 views
| Free forum by Nabble | Edit this page |