This is my output image imshow(imIDX,[]); .How can i read this image into variable and how can i make it a binary image?

figure
imshow(imIDX,[]),title('Image after applying k-means Clustering Algorithm');

How can i read the image into a single variable ? how can i make it a binary image?

[Stephen23]

imIDX is already a variable which you already have and use. imIDX is a single variable which contains the image data. Why not use imIDX?

[sam CP]

Okay..Then can i display the image by using imshow(imIDX); instead of imshow(imIDX,[]);

Answer 1

[Image Analyst]

imIDX is already a variable. To create a binary (logical) image from it, or a processed/filtered version of it, you can threshold to get values above or below some special value

binaryImage = imIDX < someThreshold;

[sam CP]

I have applied the above syntax for conversion into binary image (using some specific thrshold).but i didn't get a proper output..hereby i will attach my coding and the two sample input images

[Stephen23]

"i didn't get a proper output."

What exactly is a "proper output"?

[Image Analyst]

Man, this kmeans stuff is getting out of hand. It's getting almost as bad as people who think that edge detection is the necessary first step to any image processing.

Where did you hear that kmeans was a good way to find tumors? It's not. If you use kmeans to find clusters, it will find the number clusters you tell it to regardless of whether there is a tumor there or not. Is that what you want? Of course not. You'd be better off using a fixed threshold.

[sam CP]

Actually this tumor detection is a two step procedure..first i will get a ouput image with tumor after kmeans..Sometimes it will detect the tumor with sorrounding healthy tissue..Then HCSD method is used to remove such healthy tissue and seperate thee tumor section alone..

I got this idea from a IEEE paper. The tumor section is segmented in this ..check this code

[Image Analyst]

If you use kmeans, you're finding different gray level ranges all the time. So if you stripped off the skull and ignored the pure black mask, then you have normal gray and black in the brain, and possibly a tumor, but possibly not. So if you used a k of 3 and there was a big tumor, it could find it. However if there was no tumor, it will still find 3 classes. But unlike before, one of those classes will NOT be a tumor - it will be normal tissue. So now you're going to have to examine that class and see if that class is a real tumor or just normal tissue. For example, you could look at the absolute gray level and see if it's white or gray. Well, if you're going to do that, then why not just do that from the start and not do kmeans()? That said, I haven't read the paper and perhaps kmeans plays some role in the complete tumor segmentation process, I just don't know what it could be and still be robust enough to handle all cases from 0% tumor to 100% tumor.

http://matlab.wikia.com/wiki/FAQ#Can_you_program_up_the_algorithm_in_this_article_for_me_and_explain_it_to_me.3F