[Insight-users] Surface smoothing of labels

[Bill Lorensen]

Here is a VK example that generates a surface from labels using
Discrete Marching VUbes and then jointly smooths the resulting
surfaces. Surfaces that are touching before smoothing will continue to
touch afterwards.
http://www.vtk.org/Wiki/VTK/Examples/Cxx/Medical/GenerateModelsFromLabels

Bill

On Sat, Jan 19, 2013 at 5:34 AM, Richard Beare <richard.beare at gmail.com> wrote:
> You need to think quite carefully about what you want to achieve with
> the smoothing. Any linear-ish smoother applied to objects individually
> is likely to lead to them overlapping. This may be OK if you are only
> after visualization and/or are happy representing each image in a
> separate image.
>
> If you want to do something that keeps the properties of a label image
> then a morphological opening is a simple choice that won't introduce
> overlap. You can pull out label objects perform an opening on each
> one, put them back together. There are ways of operating directly on
> the label set quite efficiently -
> "Parallel Algorithms via Scaled Paraboloid Structuring Functions for
> Spatially-Variant and Label-Set Dilations and Erosions", Beare and
> Jackway, DICTA 2011. The code is ITK-based, but I haven't got around
> to packaging it for release, let me know if it sounds interesting
> enough to make the effort.
>
> Your other option is to smooth first - don't be tempted by using a
> chamfer (approximate) distance transform - use an exact one and smooth
> it. Also, as distance transforms are typically used to split cells, it
> suggests that you have some segmentation process prior to this. If
> this is the case you may benefit from smoothing the raw image further
> prior to segmentation.
>
> On Sat, Jan 19, 2013 at 2:51 PM, Thomas Seidel <seidel at cvrti.utah.edu> wrote:
>> Hi,
>>
>> I have segmented cells in tissue sections using a morphological watershed.
>> So the output is a label image.
>> The segmentation result is not bad. However, the surfaces of the cells seem
>> to have disturbances and some tiny artifacts (probably due to the distance
>> map which I used as gradient image).
>>
>> So I am looking for a method to smooth the surfaces of the cells without
>> loosing to much detail. I found the  itkAntiAliasBinaryImageFilter, but it
>> requires a binary image as input. I cannot reduce the label image to a
>> binary image because sometimes the cells have no background between them,
>> which means that I cannot separate them again.
>>
>> Is there any surface smoothing filter which operates on a label image?
>> How can I transform a label image into an input image for a level set
>> filter?
>>
>> Thanks,
>> Thomas
>>
>> _____________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Kitware offers ITK Training Courses, for more information visit:
>> http://www.kitware.com/products/protraining.php
>>
>> Please keep messages on-topic and check the ITK FAQ at:
>> http://www.itk.org/Wiki/ITK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.itk.org/mailman/listinfo/insight-users
>>
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://www.kitware.com/products/protraining.php
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-users



-- 
Unpaid intern in BillsBasement at noware dot com