Morphological closing to crop an image

Hi,

I'm trying to use the image processing toolbox to remove extraneous pixels from an image and I'm running into trouble. Basically I have a painmap (attached), which is a blank sheet of paper with a black body template on it, and a patient can draw in red marker over the areas of the body that cause them pain. However, often patients will draw outside of the lines and these are the pixels I'd like to crop out: everything outside of the black body outline. A friend suggested I use morphological closing to accomplish this but I'm not sure how to go about doing this. If anyone has any ideas they would be much appreciated.

Thanks, Matt

Answer 1

[Image Analyst]

No that does not sound right. You'd use masking, assuming you have the original body outline. Closing will smooth out the white parts of the image. You didn't attach anything yet. Why don't you attach and image of the blank sheet, an image where the patient scribbled on it, and what you'd like that image to look like after processing?

[Matthew Alston]

Oops I forgot to actually attach the image, they are both attached here. So image1 is the template given out to a patient, and image2 is after they've drawn on it. I'd like all of the red outside of the body template in image2 to be cropped out because only the shaded pixels within the body are relevant to me. Does this make it more clear?

[Image Analyst]

There are a variety of ways to do it, from simple to more robust. For simple starters, extract the different color channels

% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);

and threshold them to get the black pixels only,

mask = redChannel < 188;
mask = imfill(mask, 'holes');

and create a mask from the black pixels. Then mask the color channels and rebuild the image.

% Mask the image so that it's pure white outside the black body outline.
redChannel(~mask) = 255;
greenChannel(~mask) = 255;
blueChannel(~mask) = 255;
rgbFixed = cat(3, redChannel, greenChannel, blueChannel);

That's untested so you may have to tweak some code. If you get that working, there are some ways to make it a lot more robust. Ask if you get to that point.

[Matthew Alston]

Great, thanks! It could be that I'm doing something wrong but your code removed the red entirely from the picture (with some residue left behind). Is there a way to keep everything inside of the body?

[Image Analyst]

Make it easy for me and attach your code.

[Matthew Alston]

Here it is:

>> rgbImage = imread ('image2.jpg');

>> imshow (rgbImage);

Warning: Image is too big to fit on screen; displaying at 17% > In imuitools/private/initSize at 71 In imshow at 282

>> redChannel = rgbImage (:, :, 1);

>> greenChannel = rgbImage (:, :, 2);

>> blueChannel = rgbImage (:, :, 3);

>> mask = redChannel < 188;

>> mask = imfill (mask, 'holes');

>> redChannel (~mask) = 255;

>> greenChannel (~mask) = 255;

>> blueChannel (~mask) = 255;

>> rgbFixed = cat(3, redChannel, greenChannel, blueChannel);

>> imshow (rgbFixed);

Warning: Image is too big to fit on screen; displaying at 17% > In imuitools/private/initSize at 71 In imshow at 282

Would it be easier to attach as a .m file?

[Image Analyst]

Yes

[Matthew Alston]

Here it is:

[Image Analyst]

Huh???? That's not the right code. That code is for registering an image.

[Matthew Alston]

Oops! Sorry got them confused, this should be it:

[Image Analyst]

The problem was that the outline was not completely closed - it had gaps so the body did not fill in. I used imclose to fill the gaps. See attached m-file.

[Matthew Alston]

Ahh thank you so much, this is perfect.