[ITK] [ITK-users] Texture pixel by pixel

[Timothee Evain]

Well, if you do like in the example, you will have as many images as features you have.
Of course you can combine them. But I've never work with nifti images, so I can't tell you if you will be able to save it.
But since you could consider such a features image as a multi-components image (like RGB one), I think you could maybe find help on writing RGB nifti images?

Tim

PS: Try to keep posting to the ITK list for future readers :) 

----- Mail original -----
De: "Francisco Lopez de la Franca" <franciscolopezdelafranca at gmail.com>
À: "Timothee Evain" <tevain at telecom-paristech.fr>
Envoyé: Lundi 21 Septembre 2015 14:16:00
Objet: Re: [ITK] [ITK-users] Texture pixel by pixel

Thank you very much Tim.
I think the same, that there is no much sense on calculating the texture
feature per pixel, but I think the ITK filters have to do it (of course, by
using a neighbourhood region as you mention), because they return the
average on each direction.
Anyway, I'll have a look at the referenced example.

>It shows how to obtain "feature image", i.e. an image where each pixel
contain the value of a feature computed on a 3x3x3 neighborhood.
This is exactly what I need.

One more question, if I store the texture features per pixel, I mean, I'd
have a vector image (the pixel type as an array), could I save this kind of
feature image as a NIFTI image?

Thanks again.



2015-09-21 14:07 GMT+02:00 Timothee Evain <tevain at telecom-paristech.fr>:

> Hello Francisco,
>
> I'm not sure about what you are trying to do, but computing a feature on a
> single pixel does not make much sense to me.
> When you are trying to characterize a texture, you rely on the fact there
> should be a heterogeneity in an area. The feature computing try to
> highlight such differences (heterogeneity/homogeneity, constrast,
> symmetry...)
> Nevertheless, you may be interested by this example :
> http://itk.org/Wiki/ITK/Examples/Statistics/TextureFeatures
> It shows how to obtain "feature image", i.e. an image where each pixel
> contain the value of a feature computed on a 3x3x3 neighborhood.
> You just have to replace the ScalarImageToCooccurrenceMatrixFilter &
> HistogramToTextureFeaturesFilter by your ScalarImageToTextureFeaturesFilter
> (which is just the fusion of the two).
>
> About your concern for computation time, I'm afraid there is no other easy
> way than iterating a mask region. You can tune the neighborhood size to get
> acceptable results in balance with speed.
>
> HTH
>
> Tim
>
> ----- Mail original -----
> De: "Francisco Lopez de la Franca" <franciscolopezdelafranca at gmail.com>
> À: "Insight Users" <insight-users at itk.org>
> Envoyé: Lundi 21 Septembre 2015 13:15:19
> Objet: [ITK] [ITK-users] Texture pixel by pixel
>
> Hi,
> I've used the itk::Statistics::ScalarImageToTextureFeaturesFilter class
> with the default parameters and everything was OK.
>
> But now, I have to calculate the texture features of a 2D scalar image
> pixel by pixel, I mean, iterate over the whole image and calculate the
> texture features of every pixel.
>
> I guess I've got to use each pixel as the masked region over which to
> calculate the texture features, but I think this process could be very slow
> (I have to do it with a lot of 2D images).
>
> Is there a faster way to do it or isn't it possible as I've described?
>
> Thanks a lot.
> Kind regards.
>
> _____________________________________
> 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://public.kitware.com/mailman/listinfo/insight-users
>
> _______________________________________________
> Community mailing list
> Community at itk.org
> http://public.kitware.com/mailman/listinfo/community
>
_____________________________________
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://public.kitware.com/mailman/listinfo/insight-users