How to calculate diameter of object

I have the following code to find the centroids of the particles within this image <- Hyperlink:-

I would like to calculate the diameter of each of the particles so to generate the diameter distribution of the entire image.

X=Image;
template = fspecial('LoG',[51,51],3); % produces a laplacian of Gaussian curve as a template mask use to compare all the intensities in the image
y = conv2(double(X),-template,'same'); % convolves the image with the template to find all the points which correspond to the template mask
FoVY = y(500:700,730:940);
Peaks = imregionalmax(FoVY); % finds all the local maxima
imagesc(Peaks)
LPeaks = bwlabel(Peaks);
R = regionprops(LPeaks);
FoVY = X(500:700,730:940);
imagesc(FoVY);
hold on;
for i = 1:length(R)
    plot(R(i).Centroid(1),R(i).Centroid(2),'r.');
    centroid = cat(1, R.Centroid);
end
no_of_collagen=length(centroid);

Answer 1

[Image Analyst]

You can call regionprops() and then ask for EquivDiameter. You might get batter, more comprehensive answers if you put as much effort into your question here on the Mathworks MATLAB Central as you did for some other forum.

[Ahmed Ismail]

Thanks for the response Image Analyst, I've now formatted the question as it was on the other forum.

I have already attempted regionprops() EquivDiameter: diameter = cat(1, R.EquivDiameter); within the loop

but I get a matrix with diameters of the same values.

[Image Analyst]

Well I see only one particle. Do you have other examples? Is the particle you're showing actually multiple particles all stuck together?

What does this show

[LPeaks, numberOfPeaks] = bwlabel(Peaks);
R = regionprops(LPeaks, 'Centroid', 'EquivDiameter');
allDiameters = [R.EquivDiameter]
message = sprintf('Number of peaks = %d\nCheck command window for diameters.', numberOfPeaks);
uiwait(helpdlg(message));

How many peaks does it say there are?

And why do you think that the blobs are more than 1 pixel? They won't be more than one pixel unless your peaks have flat tops (uniform areas).

[Ahmed Ismail]

Ah right, I should explained that the greyscale image in question is a series of collagen fibrils stuck together in a bundle, so generally the high intensity points are individual 'particles' (collagen fibril).

As for your enquiry, there are 85 peaks (but there actually is 84, because it counts the centroid of the entire image for some reason). The diameter matrix shows the diameters for all 85 peaks, all at 1.1284 (apart from the centre of the image centroid, of course).

I must admit that finding the diameter of each individual fibril is highly optimistic, as you said, in the image, each fibril seem to have several grades of greyscale, which is the challenge here.

EDIT: I just noticed too, that in the command window the diameter is a value of 0.0011 for all, whilst it is 1.1284 in the workspace, under the same variable: allDiameter

[Image Analyst]

allDiameters should be an array of 84 or 85 elements. Why isn't it?

[Ahmed Ismail]

it is, a 1x85 array, however all the values are the same

[Image Analyst]

I tried your code but ran into some problems. I tried to fix them but ran into a problem with the cropping, so it's clear the posted image is not the same size as what you were using with your code.

X = imread('segmented_collagen.png');
X = double(rgb2gray(X));
imshow(X, []);
template = fspecial('LoG',[51,51],3); % produces a laplacian of Gaussian curve as a template mask use to compare all the intensities in the image
y = conv2(double(X),-template,'same'); % convolves the image with the template to find all the points which correspond to the template mask
FoVY = y(500:700,730:940);
Peaks = imregionalmax(FoVY); % finds all the local maxima
imagesc(Peaks)
LPeaks = bwlabel(Peaks);
R = regionprops(LPeaks);
FoVY = X(500:700,730:940);
imagesc(FoVY);
hold on;
for i = 1:length(R)
    plot(R(i).Centroid(1),R(i).Centroid(2),'r.');
    centroid = cat(1, R.Centroid);
end
no_of_collagen=length(centroid);

Please post the image you actually used and the code you actually used so I can try it.

[Ahmed Ismail]

Ah yes, the image in the question is the cropped version, the result of FoVY = X(500:700,730:940);, so for the sake of troubleshooting those line can be ignored, and Peaks = imregionalmax(FoVY), replaced with Peaks = imregionalmax(X)

I will post the original image as soon as I have access.

[Ahmed Ismail]

The entire image can found at: http://imgur.com/kL0LCZz As for the code, it should be as stated in the question.

[Image Analyst]

I tried your code and got it working. All your areas are 1 pixel - not surprising.

clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;  % Erase all existing variables.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;

grayImage = imread('segmented_collagen.png');
[rows, columns, numberOfColorChannels] = size(grayImage);
if numberOfColorChannels > 1
  grayImage = double(rgb2gray(grayImage));
else
  grayImage = double(grayImage);
end
subplot(2,3,1);
imshow(grayImage, []);
title('X', 'FontSize', 20);
% Get rid of white surround
grayImage = imclearborder(grayImage);
template = fspecial('LoG',[51,51],3); % produces a laplacian of Gaussian curve as a template mask use to compare all the intensities in the image
y = conv2(double(grayImage),-template,'same'); % convolves the image with the template to find all the points which correspond to the template mask
% FoVY = y(500:700,730:940);
FoVY = y;
subplot(2,3,2);
imshow(FoVY, []);
title('FoVY', 'FontSize', 20);
Peaks = imregionalmax(FoVY); % finds all the local maxima
% Get rid of white surround
Peaks = imclearborder(Peaks);
subplot(2,3,3);
imagesc(Peaks);
title('Peaks', 'FontSize', 20);
[LPeaks, numberOfPeaks] = bwlabel(Peaks);
R = regionprops(LPeaks, 'Centroid', 'EquivDiameter', 'Area');
allDiameters = sort([R.EquivDiameter], 'descend')
allAreas = sort([R.Area], 'descend')
% Show a histogram of the diameters
subplot(2,3,4);
histogram(allDiameters);
grid on;
title('Histogram of Diameters', 'FontSize', 20);
% FoVY = X(500:700,730:940);
FoVY = grayImage;
subplot(2,3,5);
imagesc(FoVY);
hold on;
for i = 1:length(R)
    plot(R(i).Centroid(1),R(i).Centroid(2),'r.');
    centroid = cat(1, R.Centroid);
end
title('FoVY', 'FontSize', 20);

no_of_collagen=length(centroid);

% Set up figure properties:
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0 0 1 1]);
% Get rid of tool bar and pulldown menus that are along top of figure.
set(gcf, 'Toolbar', 'none', 'Menu', 'none');
% Give a name to the title bar.
set(gcf, 'Name', 'Demo by ImageAnalyst', 'NumberTitle', 'Off') 

message = sprintf('Number of peaks = %d\nCheck command window for diameters.', numberOfPeaks);
uiwait(helpdlg(message));

I had trouble downloading your original image from imgur. You should have inserted it here where it would have been easier for us.

[Ahmed Ismail]

Here is the original, is this what you meant by inserting? As for your results, would this suggest that the collagen are unsegmentable? What I mean is that it cannot distinguish when a collagen begins and where it ends.

[Image Analyst]

It would be difficult, even manually with imfreehand. I can't really tell where one starts and the other stops. Why isn't the count enough? If you know the total area, and the count, you can get the average area of the collagen.

[Ahmed Ismail]

That is true, I guess with some images you have to be realistic with what you can extract. Thanks for the help.