masure width of binary image object

Hi,

i got an binary image:

The problem is, when i use the regionprops function, the width and height of the yellow object are not correct,see red box.

How can i find the real width and hight of the object?

Edit:

What i want to do is to eliminate regions which are very small. As you can see the yellow regions is small but with the boundingbox function it is just giving the widht and height as seen in the red box. So this region won't be eliminated. So how can i get the real width of the yellow region which is not based on the outside of the region.

[DGM]

You didn't include the image or what you did, so I can only assume that you found the 'boundingbox' property and plotted the position with a rectangle. It certainly looks like that's what you did, and it looks correct.

What is the image?

What did you do?

Why do you think it's wrong?

[soepblik]

hi DGM,

Thanks for your reply. yes i did the boundingbox and plotted it with a rectangle.

See my edit on the original message for more information.

[Image Analyst]

@soepblik, see my answer below. See how my tortuosity demo computes the mean width of the snake-like blob, similar to the shape you have. It should work with your image.

Answer 1

[DGM]

Normally, if you're trying to remove small objects from a binary image, you do it by area.

Toward that end, bwareaopen() can be used to remove objects below a specified area.

Alternatively, bwareafilt() can be used do things such as removing all but the largest object.

A = imread('blobs.png');
imshow(A)

B = bwareaopen(A,100); % get rid of blobs with <100px area
clc; imshow(B)

C = bwareafilt(A,1); % get rid of everything but the largest blob
clc; imshow(C)

If for some other reason you need to find the width of a line, you can invert the image and use bwdist() to find the distance to the background. The maximum value in the blob region will be its maximum width.

[soepblik]

thanks for your reply! the thing is, the region can be big in pixels but still very thin. so i am going to see what i can do with bwdist()

[DGM]

This is one way to find mean "thickness" of objects. It's not exact, but perhaps close enough.

A = imread('blobs.png');

D = bwdist(~A); % get distance map
S = bwmorph(A,'skel',inf); % skeletonize image

[L N] = bwlabel(A);
meanwidths = zeros(N,1);
% remove endpoints until only two are left per blob
for blob = 1:N
    thisskel = S & (L == blob);
    S(thisskel) = false;
    while true
        e = bwmorph(thisskel,'endpoints');
        if nnz(e)>2
            thisskel = thisskel & ~e;
        else
            break;
        end
    end
    S = S | thisskel; % replace with pruned skeleton
    thisMWL = thisskel.*D; % extract locus of maximal width
    meanwidths(blob) = mean(thisMWL(thisMWL~=0)); % find mean width
end

% this is the approximate midline of the region
% the distance to the bg is sampled along these lines and averaged
imshow(S)

meanwidths % average thickness for each object

meanwidths = 28×1

1.0000 4.9016 1.0000 1.0000 1.0000 1.1381 2.9082 1.0000 1.0000 1.0000

Alternatively, if you want to get rid of things that aren't very "solid", you might look at the 'extent' or 'solidity' properties (regionprops()) instead of just area. You could even use bwpropfilt() to get rid of blobs accordingly.

A = imread('blobs.png');

% get rid of long, skinny things
B = bwpropfilt(A,'eccentricity',[0 0.99]);
% get rid of things with low solidity
B = bwpropfilt(B,'solidity',[0.5 1]);

imshow(B)

Answer 2

[Image Analyst]

See my attached demo on tortuosity. It computes mean with of a blob. Adapt as needed.