CODEWITHSUNDEEP
matlabimage-processinggeometry
Hadi
Hadi

Reputation: 37313

Binary image circle fitting

I a new to MATLAB and i am trying to apply circle fitting technique to the following image.

Image

enter image description here

Expected Output

enter image description here

I used the following code with no luck:

[centers, radii] = imfindcircles(W,[30 65],'ObjectPolarity','bright');
figure,imshow(W);% title('Sobel gradient');
viscircles(centers,radii,'Color','b');

Note that the circle size can vary between picture

Any suggestions.

Upvotes: 1

Views: 622

Answers (2)

Leander Moesinger
Leander Moesinger

Reputation: 2462

So close! You just have to adjust the sensitivity a bit. It ranges from 0 to 1, the higher the more circles are found. By default it is 0.85. Also had to adjust the radius a bit:

imshow(W) % assuming W is logical
[centers,radii] = imfindcircles(W,[40 60],'ObjectPolarity','bright','Sensitivity',0.925)
viscircles(centers, radii,'EdgeColor','b');

I don't think it's possible to find the lower circle without also getting some false positives first - ultimately it's not very round.

resultimage

Upvotes: 4

Tommaso Belluzzo
Tommaso Belluzzo

Reputation: 23675

I tried to play a little bit with this image despite not being a good digital images analyst. I found the following approach much more flexible:

img = imread('1GBtO.png');
img = im2bw(img);

rp = regionprops('table',img,'Centroid','Eccentricity','EquivDiameter','MajorAxisLength','MinorAxisLength');
rp.Radii = mean([rp.MajorAxisLength rp.MinorAxisLength],2) / 2;

rp([rp.Eccentricity] > 0.75,:) = [];
rp(rp.EquivDiameter < 10 | rp.EquivDiameter > 30,:) = [];

imshow(img);
hold on;
viscircles(rp.Centroid,rp.Radii);
hold off;

The output is far from being the desired one, but I feel it's an improvement over your first trial:

Output

But I think that you can obtain the desired result by finding the right algorithm to adjust the radii and by tweaking the criterions to exclude unwanted regions.

Upvotes: 1

Related Questions