Detecting and Counting Objects with Circular Features
This example shows how you can use imfindcircles together with removeoverlap function (can be found in the file exchange), for counting fungi spores, which has various elliptical like shapes.
As you can see, the spores have different shapes and sizes, some overlap objects from other planes of the imaged sample. This is a common microscopy problem in biology.
First, try to find circles in the image using imfindcircles. For estimating the radius range of our objects we can use imdisline:
Using a radius range between 12 to 30 pixel and visualizing the results with viscircles:
[centers, radii] = imfindcircles(image,[12 30]); close all;figure; imshow(image); viscircles(centers, radii,'EdgeColor','b');
Lets increase the Sensitivity factor (the default is 0.85) and use a low static Edge Gradient Threshold instead of the default graytreshold.
[centers, radii] = imfindcircles(image,[12 30],'Sensitivity',0.92,'Edge',0.03); close all;figure; imshow(image); viscircles(centers, radii,'EdgeColor','b');
Now, it seems we detected more circles than spores, mostly because of overlapping circles. Using removeoverlap function we can remove the overlapping circles, or allow an overlap of circle pair up to some tolerance, e.g: 5 pixels overlap.
[centersNew,radiiNew]=RemoveOverLap(centers,radii,5,1); close all;figure; imshow(image); viscircles(centersNew, radiiNew,'EdgeColor','b');
We got a relatively good detection for the number of spores, finally we can count the number of circles.
length(centersNew) ans = 94
So, we counted 94 spores!