Diameter of Almost circular objects and degree of circularity

Hello, I am trying to measure the diameter of central white spot and the dark area.

I understand they perfect circles but I'd like to get a close estimation and also a measure of circularity (I believe by a ratio of perimeter^2/area).

I thought imfindcircles would work, I have tried with both a binarised image and just the raw image (12 bit Tiff)

The size of my image is 300x300 pixels.

This was my attempt, - but its failing to find any circles.

            IM=getimage(app.UIAxes2);
            figure
            ax1=subplot(1,3,1);
            myImagesc(app,ax1,IM);      %My version of ImageSc (i.e. force colormap gray)

            %Binarize if required
            mn=mean2(IM)
            th=mn*0.5

%             IM(IM<th)=0;    
%             IM(IM>=th)=1;
            ax2=subplot(1,3,2);
            myImagesc(app,ax2,IM);

            %Find Circles
            [centers, radii, metric] = imfindcircles(IM,[15 400],'ObjectPolarity','dark')   %I've also tried "bright"         
            viscircles(ax2,centers, radii,'EdgeColor','b');

Answer 1

[Image Analyst]

Just threshold, call imclearborder, and then regionprops, asking for EquivDiameter. Let me know if you can't figure it out.

[Jason]

Wow that imclearborder is awesome! Not seen it before. Im also playing about with filling the central bright circular part in using (after creating a binary image bw2 and inverting it so that smaller central spot is now black

BW2 = imfill(bw2,'holes');

But I dont always have that central bright spot, is there a way to not do the imfill if its not there?

Thanks

[Image Analyst]

Not sure I understand. So you have an image with just white surround and a black circle in the middle, so the black is circular, not donut-shaped? Please attach one of those images. I'm not sure what you want to fill if it's just a white frame with black circle in it. If you filled the black circle, the whole image would be white.

[Jason]

sorry, hopefully this is clearer.

So "A" is my raw image. The dark circle is actually a (darkfield) image of a water drop. The white circle inside is a bubble. So my aim is to estimate the size of the drop, and where present estimate the size of the bubble.

Firstly, to get the main drop (i.e. blue circle on A), I binarise, then perform an imfill, then use regionprops.

And when a buble is present, I use the Imclearborder followed by an regionprops.

My issue is that I won't always have a bubble, i.e the bright spot shown by the green cross. So any imfill or imclearborder may mess up or crash if I just have only the larger dark circle. So I was going to put some IF statements in for the imfill and imclearbound, but Im struggling to work out what to put in the IF statement.

[Image Analyst]

So I take it that you want the white spot inside the black disc, but there may or may not be a white spot present, right?

Is there always just one bubble/drop inside the black disc? Is there a minimum size they/it will be? Like anything smaller than 100 pixels is noise? What I'd do is something like this

mask = grayImage > 128
mask = imclearborder(mask); % Get rid of surround.
mask = bwareafilt(mask, 1); % Take largest blob only.
mask = imfill(mask, 'holes'); % In case there are any black specks inside the white drop.
props = regionprops(mask, 'Centroid', 'EquivDiameter');
if ~isempty(props)
    xCenter = props.Centroid(1);
    yCenter = props.Centroid(2);
    viscircles([xCenter, yCenter], props.EquivDiameter, 'Color', 'r');
end

[Jason]

Thanks