Feature/diffusion add an-isotropic diffusion to vmtkimagesmoothing #249

kayarre · 2018-03-02T00:21:18Z

Adding additional smoothing filter to vmtkimagesmoothing using ITK an-isotropic diffusion.

The anisotropic diffusion feature includes the ability to work with 2D or 3D and on many variable types.

rlizzo · 2018-03-02T19:08:21Z

This is great Kurt! awesome job. I'm just digging into this now, so expect updates soon.

Just one issue off the bat with the branch history. it seems that this was based off of your prior PR #196, which we decided to close and integrate the changes directly into vmtksurfaceendclipper in PR #200. Basically, this means that the current history of the PR would recreate the previously proposed changes and create a new vmtkScripts/vmtksurfaceclippercenterline.py.

I'm going to attempt to fix this now so that the proposed changes are based off the current master branch, will let you know if I run into any weird issues that would better be handled by you as author of the PR.

rlizzo · 2018-03-02T19:53:02Z

Ok. So the history is kindof a mess and can't be easily fixed by me via a single merge or few cherry picks (atleast while preserving history with you as the author). I think it'll be easiest for you to handle on your end.

Following the git log here, we see that 822a3e9 is based off of f45a8fd and 70f4349, which is in turn based off of 5aaaa16 (which was current state of master roughly 5 months ago). As f45a8fd and 70f4349 were part of our rejected PR, this leaves the history significantly out of date.

You need to pull the latest upstream master branch into your fork's master. Then checkout your feature/diffusion branch and rebase it off of master. This fixed it in my tests. It's probably cleanest to then submit the rebased branch as a seperate PR so that the public histories align.

kayarre · 2018-03-02T22:07:46Z

Thanks rick, I was fuzzy braining a lot yesterday. I will fix it up.

…

-Kurt

On Fri, Mar 2, 2018 at 11:53 Rick Izzo ***@***.***> wrote: Ok. So the history is kindof a mess and can't be easily fixed by me via a single merge or few cherry picks (atleast while preserving history with you as the author). I think it'll be easiest for you to handle on your end. Following the git log here, we see that 822a3e9 is based off of f45a8fd and 70f4349, which is in turn based off of 5aaaa16 (which was current state of master roughly 5 months ago). As f45a8fd and 70f4349 were part of our rejected PR, this leaves the history significantly out of date. [image: image] <https://user-images.githubusercontent.com/17350115/36918218-31689146-1e27-11e8-8199-88eae479807e.png> You need to pull the latest upstream master branch into your fork's master. Then checkout your feature/diffusion branch and rebase it off of master. This fixed it in my tests. It's probably cleanest to then submit the rebased branch as a seperate PR so that the public histories align. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#249 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AInOVYT5PGRHZ64eYzd4vf-jf4_-5niuks5taaMfgaJpZM4SZNaF> .

kayarre · 2018-03-05T18:35:28Z

@rlizzo I am sorry about this headache I created. I was able to rebase/revert create a new commit and force it to feature/diffusion. I tried to follow your instructions while also trying to fix my remote master.

rlizzo · 2018-03-05T22:27:44Z

vtkVmtk/Segmentation/vtkvmtkAnisotropicDiffusionImageFilter.cxx

+  }
+
+
+//   using PixelType = float;


Remove unused code, or state why it is temporarily commented out

rlizzo · 2018-03-05T22:27:44Z

vtkVmtk/Segmentation/vtkvmtkAnisotropicDiffusionImageFilter.cxx

+      break;
+    }
+
+    case VTK_CHAR:


Shouldn't a char type fail every time. Not sure that the processing algorithm would be able to handle chars

This portion was from something similar in ITK. I happy to remove the char type. I would assume that in most cases that makes sense. I don't have a good idea of which types are or should be handled.

Ok. Yeah, from the documentation (https://itk.org/Doxygen/html/classitk_1_1GradientAnisotropicDiffusionImageFilter.html) it seems that only scalar image types are supported. Would probably be best to just remove it to prevent confusion in the future.

What's the definition of a scalar type?
I tried this in slicer where I cast an image to char and filtered with anisotropic diffusion and it worked without a problem. I believe there is a separate way to check whether the input image is a scalar image volume which i think addresses the scalar type issue. otherwise there is no reason the filter can't operate on each type?

is the concern here that we need to test each data type? or something else?
in all likelihood it's not an often used use case, but I have a hard time not being thorough.

I was a bit confused it seems. I've been doing a bit of Java and got confused between the C++ chars and Java chars (which are essentially unicode characters). The VTK_CHAR is just an 8 bit value (integers between 0 and 255), which seems like it's fine for this application.

My understanding is that a scalar type (used in this context) is any numeric value that VTK can handle. In any case, it's not an issue, sorry about the confusion!

rlizzo · 2018-03-05T22:27:44Z

vmtkScripts/vmtkimagesmoothing.py

        self.StandardDeviation = 1.0
        self.RadiusFactor = 5.0
        self.Dimensionality = 3
+        self.Conductance = 1.0
+        self.NumberOfIterations = 5
+        self.TimeStep = 0.0625


Not sure if this timestep is the best default. For the standard image we use in our test cases, I recieved the following output while using the default parameters.

... Automatic piping vmtkimagesmoothing Image = vmtkimagereader-0.Image Parsing options vmtkimagesmoothing Method = anisotropic Explicit piping vmtkimagesmoothing Input vmtkimagesmoothing members: Id = 0 Disabled = 0 Image = vtkImageData ImageInputFileName = Method = anisotropic StandardDeviation = 1.0 RadiusFactor = 5.0 Dimensionality = 3 Conductance = 1.0 NumberOfIterations = 5 TimeStep = 0.0625 ImageOutputFileName = Executing vmtkimagesmoothing ... WARNING: In /Users/rick/anaconda3/envs/vmtkbuild/include/ITK-4.13/itkAnisotropicDiffusionImageFilter.hxx, line 82 GradientAnisotropicDiffusionImageFilter (0x7fc46cf20600): Anisotropic diffusion unstable time step: 0.0625 Stable time step for this image must be smaller than 0.0549316 WARNING: In /Users/rick/anaconda3/envs/vmtkbuild/include/ITK-4.13/itkAnisotropicDiffusionImageFilter.hxx, line 82 GradientAnisotropicDiffusionImageFilter (0x7fc46cf20600): Anisotropic diffusion unstable time step: 0.0625 Stable time step for this image must be smaller than 0.0549316 WARNING: In /Users/rick/anaconda3/envs/vmtkbuild/include/ITK-4.13/itkAnisotropicDiffusionImageFilter.hxx, line 82 GradientAnisotropicDiffusionImageFilter (0x7fc46cf20600): Anisotropic diffusion unstable time step: 0.0625 Stable time step for this image must be smaller than 0.0549316 WARNING: In /Users/rick/anaconda3/envs/vmtkbuild/include/ITK-4.13/itkAnisotropicDiffusionImageFilter.hxx, line 82 GradientAnisotropicDiffusionImageFilter (0x7fc46cf20600): Anisotropic diffusion unstable time step: 0.0625 Stable time step for this image must be smaller than 0.0549316 WARNING: In /Users/rick/anaconda3/envs/vmtkbuild/include/ITK-4.13/itkAnisotropicDiffusionImageFilter.hxx, line 82 GradientAnisotropicDiffusionImageFilter (0x7fc46cf20600): Anisotropic diffusion unstable time step: 0.0625 Stable time step for this image must be smaller than 0.0549316 Done executing vmtkimagesmoothing. Output vmtkimagesmoothing members: Id = 0 Image = vtkImageData ...

The time step comes from (PixelSpacing)/2^{N+1}, so for 1.0/2^{4}=0.0625
This is the default in slicer, which I have used (obviously this warning message is hidden). I was thrown off at first, but there is really no way to know apriori what the user wants. We could have the script check for the largest dimension and calculate the minimum time step as the default and only override is the user specifies one?
the challenge here is that obfuscates the interface a little because the script will print out the default time step so another message would be required to inform the user that the timestep is now different.

Ok. thanks for the clarification!

Since I'm assuming that we want to use the correct value (calculated from the equation) for most people, in most cases, what if we included a method to auto-calculate-timestep, along with an option which defaulted to use that? We could then return the used timestep as an output member to the script, ensuring that the user could backtrack and figure out which value was used (sort of like how isosurface value can be returned in http://www.vmtk.org/vmtkscripts/vmtkimageinitialization.html).

If the auto-calculate-timestep option was enabled, it would just overwrite whatever value of self.TimeStep was provided in the __init__ method. If it was disabled, we would use the self.TimeStep value provided by default, or override it automatically with a user defined value if they decided they wanted to tweak it manually.

I think that would give us both customizability, and standardized logging in one go?

fantastic idea, will implement this. I also am going to multiply the calculated time step by a factor close to but less than 1 so it doesn't throw the warning.

rlizzo · 2018-03-05T22:33:32Z

Hey kurt, thanks for fixing this up. The history seems to be in order, and this is a great addition! really nice work!

I added a couple comments in the review above, but in general things seem to work well. After we discuss the details above, we just need to create a few tests for the new parameters. I wrote up a short bit recently on how to contribute tests. You can find it here. Let me know if you have any questions or issues with how the testing process should work, would love to get your feedback on how to improve the documentation.

Thanks again! let me know your thoughts!

rlizzo · 2018-03-05T22:38:31Z

vtkVmtk/Segmentation/vtkvmtkAnisotropicDiffusionImageFilter.cxx

+Module:    $RCSfile: vtkvmtkAnisotropicDiffusionImageFilter.cxx,v $
+Language:  C++
+Date:      $Date: 2006/04/06 16:48:25 $
+Version:   $Revision: 1.1 $


Update Date to current and Version to 1.4

rlizzo · 2018-03-05T22:38:54Z

vtkVmtk/Segmentation/vtkvmtkAnisotropicDiffusionImageFilter.cxx

+Date:      $Date: 2006/04/06 16:48:25 $
+Version:   $Revision: 1.1 $
+
+  Copyright (c) Luca Antiga, David Steinman. All rights reserved.


Copyright (c) Luca Antiga. Remove David Steinman

rlizzo · 2018-03-05T22:39:48Z

vtkVmtk/Segmentation/vtkvmtkAnisotropicDiffusionImageFilter.cxx

+     This software is distributed WITHOUT ANY WARRANTY; without even 
+     the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR 
+     PURPOSE.  See the above copyright notices for more information.
+


Add note:

Note: this class was contributed by Kurt Sansom (affiliation or github username if desired)

rlizzo · 2018-03-05T22:40:48Z

vtkVmtk/Segmentation/vtkvmtkAnisotropicDiffusionImageFilter.h

+Module:    $RCSfile: vtkvmtkAnisotropicDiffusionImageFilter.h,v $
+Language:  C++
+Date:      $Date: 2006/04/06 16:48:25 $
+Version:   $Revision: 1.2 $


Update date and version number to 1.4.

add commented note:

Note: this class was contributed by Kurt Sansom (affiliation or github username if desired)

kayarre · 2018-03-05T23:17:40Z

@rlizzo thank you for the input. I will take a look at the testing later this week.

…cted

rlizzo · 2018-03-13T05:18:58Z

I made a few changes and added the test cases. Merging this now! Thanks for the contribution @kayarre !

kayarre mentioned this pull request Mar 2, 2018

frangi vesselness filter hangs #243

Open

rlizzo self-assigned this Mar 2, 2018

add anisotropic image filter to imagesmoothing script

78c8d35

kayarre force-pushed the feature/diffusion branch from e7c6a66 to 78c8d35 Compare March 5, 2018 18:14

update the timestep defaults in vmtkimagesmoothing

bc1301e

rlizzo reviewed Mar 5, 2018

View reviewed changes

address comments of copyright, contribution, and commented code

90af9c5

kayarre and others added 3 commits March 6, 2018 09:34

add automatic timestep calculation as default

fde19f2

cleanup and comment code. add checks to ensure correct method is sele…

8c2d361

…cted

updated vmtk-test-data submodule pointer

d0e92d3

rlizzo merged commit b4f2f61 into vmtk:master Mar 13, 2018

kayarre deleted the feature/diffusion branch December 7, 2021 02:33

Feature/diffusion add an-isotropic diffusion to vmtkimagesmoothing #249

Feature/diffusion add an-isotropic diffusion to vmtkimagesmoothing #249

kayarre commented Mar 2, 2018

rlizzo commented Mar 2, 2018

rlizzo commented Mar 2, 2018

kayarre commented Mar 2, 2018 via email

kayarre commented Mar 5, 2018

rlizzo Mar 5, 2018

rlizzo Mar 5, 2018

kayarre Mar 5, 2018

rlizzo Mar 5, 2018

kayarre Mar 6, 2018

rlizzo Mar 13, 2018

rlizzo Mar 5, 2018

kayarre Mar 5, 2018

rlizzo Mar 5, 2018

kayarre Mar 6, 2018 •

edited

rlizzo commented Mar 5, 2018

rlizzo Mar 5, 2018

rlizzo Mar 5, 2018

rlizzo Mar 5, 2018

rlizzo Mar 5, 2018

kayarre commented Mar 5, 2018 •

edited

rlizzo commented Mar 13, 2018

Feature/diffusion add an-isotropic diffusion to vmtkimagesmoothing #249

Feature/diffusion add an-isotropic diffusion to vmtkimagesmoothing #249

Conversation

kayarre commented Mar 2, 2018

rlizzo commented Mar 2, 2018

rlizzo commented Mar 2, 2018

kayarre commented Mar 2, 2018 via email

kayarre commented Mar 5, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kayarre Mar 6, 2018 • edited

Choose a reason for hiding this comment

rlizzo commented Mar 5, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kayarre commented Mar 5, 2018 • edited

rlizzo commented Mar 13, 2018

kayarre Mar 6, 2018 •

edited

kayarre commented Mar 5, 2018 •

edited