How to extract the main square, image window of the ultrasound image?

Hi

I need some help.

How do I extract/crop the main square of the ultrasound image. The image window that shows the valuable information and discard all the rest black and annotations.

Looking forward to your valuable help

Thanks

Answer 1

[Constantino Carlos Reyes-Aldasoro]

Ok, the issue here is the colour maps and the conversion of uint8 to double. If you convert to double the colours are always between 0 and 1, so you would need to divide by the maximum value (in this case 255)

imagesc(double(A6__14_9N_).*(repmat(1-(background),[1 1 3]))/255)

OR you can convert to uint8 the background:

imagesc(A6__14_9N_.*(repmat(1-uint8(background),[1 1 3])))

Both should work.

[Stelios Fanourakis]

Yes they work. Thank you very much. And how do I remove the black background and keep only the gray square?

[Constantino Carlos Reyes-Aldasoro]

Ok, that is a bit different. You basically need to find the limits of rows and columns. Find the complement of the background

(1-background) 

and sum over the columns and rows

sum(   ,1)
sum(   ,2)

That would generate 1D vectors where the foreground exist. The use the function "find" and select the first non-zero element (that will give you the first row/column) and the last (final row/column) and with that you can crop into a new matrix.

Answer 2

[Constantino Carlos Reyes-Aldasoro]

Hard to answer without knowing your data. Post a sample and we can try.

[Stelios Fanourakis]

See attached image. Thanks

Answer 3

[Constantino Carlos Reyes-Aldasoro]

*IF* all the images look like this, it is rather easy, EXCEPT for the fact that the bottom edge of the ultrasound is not very well defined.

The key is to find the background to determine the foreground, The background is black, so take only one channel (the image is RGB), compare against zero and that would give the background but with the annotations and bits you want to remove, so close the image with imclose with a structural element sufficiently large to cover those elements:

background =(imclose(A6__14_9N_(:,:,1)==0,ones(25)));

Then, you only need to take the complement and recover the ultrasound part

imagesc(A6__14_9N_.*(repmat(1-background,[1 1 3])))

If you compare with the original, it seems that the job is done.

[Stelios Fanourakis]

Yes exactly that. How do I crop the black area from your first image so it remains only the gray square?

Another example. Please see the one I attach after a thresholding I've done to the image. Imaging this red bounding box it would be bigger and it could capture the whole gray square and then crop it. Simple as that. Thanks once again

[Stelios Fanourakis]

I cannot make it run like you do. The background line works but the imagesc not. I convert it to double and gives me a pure black image

[Stelios Fanourakis]

Also I get a black and white blob and not the rectangular of the ultrasound image

[Stelios Fanourakis]

I think that when I convert it to double it gets black and white.

[Stelios Fanourakis]

Did you convert it from .bmp to other format?