removing flim artifacts from image

I have an MRI image in which it contains artifacts and white labels,kindly tell how to remove these labels and artifacts,I have uploaded the image

[Matt J]

I hope the MRI subject gave consent to have his identifying medical information published. Otherwise, HIPAA would probably mandate that you take that image down.

[Walter Roberson]

I removed the image link pending clarification on this issue.

[kash]

Sir there is no other way proceeding this

Answer 1

[Image Analyst]

If you want to do it automatically, if you can depend on the white letters and other annotations being 255 or 65535, then you can just erase them

maxValue = max(grayImage(:))); % most likely 255 or 65535 depending on 8 or 16 bit.
whiteStuff = grayImage == maxValue;
grayImage(whiteStuff) = 0; % Set white things to black.

Now if there are parts inside the skull that could also be the sam brightness then you'd need to use a different strategy. You'd have to find the background intensity, threshold, and find the largest blob (see my attached demo). Then invert that binary image and mask your original with it.

binaryImage = grayImage > 50; % or whatever value the brightest part of the background has.
binaryImage = imfill(binaryImage, 'holes');
largestBlob = ..... - see my demo
% Now erase
grayImage(~largestBlob) = 0;

This is probably more robust than the first method because it can handle super bright things inside the skull.

[kash]

I tried ur answer but still i get some white labels(1st menthod).I read the article published which consists of

Step 1: Read the MRI image and store it in a two dimensional matrix.

Step 2: Select the peak threshold value for removing white labels.

Step 3: Set flag value to 255.

Step 4: Select pixels whose intensity value is equal to 255.

Step 5: If the intensity value is 255 then, the flag value is set to zero and thus the labels are removed from MRI.

Step 6: Otherwise skip to the next pixel.

I tried these steps but the threshold differs,can u help me in solving these steps please

[Image Analyst]

It could be because you have a JPG image. If you do some of the letters may be blurred. You might have to drop the threshold. But, like I said, the second method I offered is more robust. What happened when you tried that method?

Can you reupload your image with the "Personally Identifiable Information" removed?

[kash]

here is the link to new image

http://imgur.com/aBcM9Ua

for the second method i could not get the link for demo,can u tell which demo i have to work on

[Image Analyst]

Yes, I think my second method of finding the largest blob would work perfectly. Can you try it? I attached a demo of finding the largest blob for you to review below.

[kash]

Thanks finally i got the answer

[kash]

One thing it work for one image and not for another Image ,why this occurs

[Image Analyst]

It would not work where the skull is not the largest blob in the image. For example if you have a slice with just the very tip of the skull but have some big text or logo in the image. You can attach the image where it doesn't work and I'll tell you why.

[kash]

this is the image

http://imgur.com/Z2qdSAq

where half part of the image is displayed and other gets deleted with the label

[Image Analyst]

Do you get the head but the M is attached to it because it's touching/connected to it?

[kash]

I got the result as

http://imgur.com/8NYcRGw

[Image Analyst]

What threshold did you pick? I'd probably pick around 25.

[kash]

for each image have i to pick different threshold for each image

[Image Analyst]

Why? Shouldn't the black background have about the same intensity in each image?

[kash]

ok thanks you

Answer 2

[Matt J]

I think you could just use H=imfreehand(...) to draw an ROI around the anatomy that you want to keep.

Then you can do

 mriImage=mriImage.*(~H.createMask);

to get rid of it.