Thanks for your replay. I will try to handle this as you suggested.
Regards,
Safaa
On Tuesday, February 4, 2014 1:58 AM, Gabriel Landini <
[hidden email]> wrote:
On Monday 03 Feb 2014 11:08:02 Safaa Al-Haj Saleh wrote:
> I am applying K-means clustering on histopathological medical images to
> classify the pixels into classes. I am doing this using K-means clustering
> plugin (which is an installed plugin). Everything is OK while clustering.
> The problem is when doing clustering many times, image pixels appear in
> different colors each time. How to make the colors of pixels the same when
> applying clustering many times?
Well you can't because those colours are just a label for a particular cluster
and the index/order of those clusters is arbitrary.
What you could do is investigate what each cluster contains, i.e. work out
what are the colours, shapes, sizes or textures of the regions indexed by each
cluster and then reason that out. For instance if you are getting a cluster
whose regions in the original all have colour close to (255,255,255) (white)
that is likely to be the empty background of the histological image, if they
are all roundish and the right size and colour, they are probably nuclei and
so on. But for this to work, your clustering result must very good and
represent the things you are expecting to extract.
Cheers
Gabriel
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html
Locked
