counting the number of objects

I have an brain image and there are three overlapped objects.The overlapped object is nails ..can i count these overlapped objects.in medical images

[Jan]

@Pat: This is your 108th question. It is time to include the required details to make it easy to answer.

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

[Pat]

I have uploaded my image

http://imgur.com/n7vGa

[Chandra Kurniawan]

Hello, Pat.

You said that the three nails were overlapped.

But, I have seen your uploaded image,

and they aren't overlapped each other.

[Pat]

THE nails are overlapped on brain image ,,not on each other,is it possible to count these nails

[Image Analyst]

What other image? You gave the non-overlapped image only, for some reason. Why did you even give that image at all (the solution is trivial)? When you are going to post the overlapped one?

[Pat]

only that image was given to me ,,,thats y i had posted it...

may i know what u mean by

No need - the patient is dead! ;-)

[Image Analyst]

If I had three nails embedded in my brain, most likely I would be dead, and a physician could instantly see how many nails had killed me, and wouldn't need an image processing program to count them. So why did you ask about overlapping nails when that's not what you have?

[Chandra Kurniawan]

Hahaha :D

I Agree with U

Answer 1

[Chandra Kurniawan]

Hello, Pat.

It is possible to count the nails.

You just need to perform some morphologichal operations, such as

erosion, area open, etc.

Then, in the next step you need to count the objects. You can use regionsprops command.

Here my demo :

Irgb = imread('n7vGa.jpg');
Igray = rgb2gray(Irgb);
Ibw = im2bw(Igray, 0.9);

se = strel('square',2);
Imorph = imerode(Ibw,se);
Iarea = bwareaopen(Imorph,130);
Idiff = Imorph - Iarea;
Ifinal = bwareaopen(Idiff,70);

stat = regionprops(Ifinal,'BoundingBox');
str = sprintf('number of detected nails : %d', length(stat));
imshow(Irgb); title(str); hold on;
for cnt = 1 : length(stat)
    bb = stat(cnt).BoundingBox;
    rectangle('position',bb,'edgecolor',rand(1,3));    
end

And the result :

Is that you meant??

[Pat]

Thanks chandra

[Pat]

Chandra please look at this

http://imgur.com/HALTv

even when there is a round object it is not detecting

[Chandra Kurniawan]

So, you need to apply a little different techniques.

Just try to modify my code.

Try to apply another morphologichal operations.

If you still difficult to understand that,

Then show me your original image.

[Pat]

chandra the same code cannot be applied for different images

[Pat]

http://imgur.com/jlnuV

in this at centre there are 4 overlapped parts ehich are brighter in colour ,,,,

[Chandra Kurniawan]

As long as the background image is same, you don't need to modify

your code.

Please upload the original version of http://imgur.com/HALTv

I'll try to find what's wrong with my code.

[Pat]

Chandra the original version is which i sent u first with 3 nails

i need some more objects as i was told by my guide,so i modified it , i need 5 objects ,so i modified like that

[Chandra Kurniawan]

Now, What can I do to help you solve your problem?

I think my code will also works on image 'http://imgur.com/HALTv'.

Coz, the missed object in this image (that contains 5 objects) is detected in previous image (that contains 3 object).

[Pat]

It detected only 4 objects

[Chandra Kurniawan]

Finally, I found the mistake.

The background image that you used are not same.

Check it carefully! The second one is lighter.

Now, just modify line

Ifinal = bwareaopen(Idiff,70);

with

Ifinal = bwareaopen(Idiff,63);

It works on both.

If in the future it doesn't work again,

Just try to setting the value like I did.

I hope this helps you.

[Image Analyst]

No need - the patient is dead! ;-)

[Pat]

Thank u chandra

Answer 2

[Image Analyst]

Probably . . . as long as they're not 100% overlapped and you can come up with some algorithm for separating them.

[Pat]

I have uploaded my image

http://imgur.com/n7vGa