Extract specific class from segmented image using fuzzy c means in MATLAB

Question

0.00/5 (No votes)

See more:

Hello, I look for the easiest way for extracting specific cluster of segmented image using fuzzy c means.

For example I divided my gray image two 6 clusters, then I just want to keep 4th cluster of the image, and change other clusters to black color.

Thank you in advance for providing the solution.

Posted 6-Oct-14 8:21am

Zeinab Shajirat

Add a Solution

2 solutions

Add a Solution

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Sergey Alexandrovich Kryukov · Answer 1 · 2014-10-06T09:18:00

Solution 1

First of all, you need to decide how to represent "this extraction". As the belonging to the cluster is fuzzy, "degree of belonging" or each point to a cluster is some value between 0 to 1, "extraction" is not a fully defined operation.

For example, you can 1) introduce some threshold dividing "extracted" and "dimmed" subset of points by this fuzzy-set "degree of belonging" value; 2) "dim" each pixel by some factor proportional to this "degree of belonging" value, so the pixel value with degree of 0 was totally black, and those with degree of belonging 1 was totally untouched; 3) formulate some other criterion for representing the pixels with the "degree of belonging" more than 0 but less than 1.

See also:

http://en.wikipedia.org/wiki/Fuzzy_clustering#Fuzzy_c-means_clustering[^],
http://home.deib.polimi.it/matteucc/Clustering/tutorial_html/cmeans.html[^].

All required data is available from Matlab fcm; please learn the documentation: http://www.mathworks.com/help/fuzzy/fcm.html[^].

—SA

Posted 6-Oct-14 9:18am

Sergey Alexandrovich Kryukov

Updated 6-Oct-14 9:20am

v2

Comments

Zeinab Shajirat 7-Oct-14 3:17am

Thank you Sergery, for your help. After of implementing fuzzy c means on my picture i will have for example 4 clusters, which values of 0.0.001, 0.125, 0.2987 and 0.5083 belong to clusters of 1 through 4 respectively. Now, i just want in one time to keep value of 0.125 and make other zero and in another time i just want to keep value of 0.2987 and change other clusters to zero (black). Thank you in advance for considering my question.

Sergey Alexandrovich Kryukov 7-Oct-14 11:10am

Not clear. (And didn't you already accept one answer formally, yours? Really? :-)
—SA

Zeinab Shajirat 7-Oct-14 11:25am

My mother tongue is not english. I did not understand your question, sorry! :)

Sergey Alexandrovich Kryukov 7-Oct-14 11:27am

Mine, too. I just said that I did not understand what is the remaining problem and how your answer post solved it.
—SA

Zeinab Shajirat 7-Oct-14 11:35am

Actually, i should write more code to solve the problem completely. But the code i provided is core for solving my problem.

Sergey Alexandrovich Kryukov 7-Oct-14 12:03pm

All right, Zeinab.
Good luck, call again.
—SA

[no name] 7-Oct-14 9:30am

Sometimes it seems hard to believe, that someone has so a broad knowledge. But it's good to know, that there is someone, who you can ask.
Kind regards, Bruno, my 5

Sergey Alexandrovich Kryukov 7-Oct-14 10:59am

Thank you, Bruno.
—SA

Zeinab Shajirat 7-Oct-14 11:21am

Thank you,
It was the first time that i ask a question in this forum. Sorry if it was not clear.

[no name] 7-Oct-14 11:30am

I think you addressed this to --SA and choose the wrong reply button ;)

Zeinab Shajirat 7-Oct-14 11:32am

I addressed it for you, too.

[no name] 7-Oct-14 11:35am

You don't have to apologize for anything. I made hundreds of mistakes here ;) I'm simply here to learn.
Regards, Bruno

Zeinab Shajirat 7-Oct-14 11:45am

Thank you Bruno, you posted an intelligent quote.

[no name] 7-Oct-14 12:01pm

Thank you for your compliment.
And please trust me from Sergey Alexandrovich Kryukov we can learn a lot. He is very direct, but you will learn to handle this ;)

Sergey Alexandrovich Kryukov 7-Oct-14 12:19pm

Thank you for your kind words, Bruno.
Thank you, Zeinab, for asking interesting questions and handing it nicely. It's a pleasure to help someone (even though it wasn't actual help in this case as you already did it before I advised) who actually can make things working.
—SA

Zeinab Shajirat 7-Oct-14 12:38pm

Thank you, They are powerful words for me from great expert like you.

[no name] 7-Oct-14 12:45pm

It is only the truth, respectively what I have experienced so far!
Kind regards, Bruno :)

Zeinab Shajirat · Answer 2 · 2014-10-07T02:45:00

Solution 2

"uv" contains unique values, "n" contains occurrences of each unique value.

[uv,~,indx] = unique (fuzzy_img);

n = accumarray(indx(:),1);

Posted 7-Oct-14 2:45am

Zeinab Shajirat

Comments

Sergey Alexandrovich Kryukov 7-Oct-14 11:03am

How can this be an answer to your question? And how could you even self-accept it formally?
—SA

Zeinab Shajirat 7-Oct-14 11:17am

I'm sorry if my question was not clear. Actually it is first time that i ask question in this website. I will try to ask clear questions in future.

Actually, i have abdominal MRI images, which i segment them by fuzzy c means into 4 clusters. I want to know that the liver will fall in which cluster. So i need know label or pixel value of each cluster, by comparing occurrences of each value i will understand which one is desired cluster.

The solution that u provided for me, was used in my code previously.
In the continue i needed the code that i found.

Zeinab Shajirat 7-Oct-14 11:19am

I reject my solution. But the solution was what i want.
Thank you

Sergey Alexandrovich Kryukov 7-Oct-14 11:22am

Then I did not understand why. Anyway, if this is the only missing piece, my congratulations!
—SA

Zeinab Shajirat 7-Oct-14 11:27am

Thank you for your time and help.

Sergey Alexandrovich Kryukov 7-Oct-14 12:28pm

You are very welcome.
Good luck, call again.
—SA

Extract specific class from segmented image using fuzzy c means in MATLAB

2 solutions

Solution 1

Solution 2

Add your solution here

Preview 0

Extract specific class from segmented image using fuzzy c means in MATLAB

2 solutions

Solution 1

Solution 2

Add your solution here

Preview 0

Existing Members

...or Join us