Average Hue for Multiple ROIs

Hi,

I'm learning MATLAB on the fly within my job, so am very much a novice. I have an image of 9x circles in a 3x3 fashion. So far I have created a binary mask and have applied it to disregard the background and to only include these circles. From here I converted my new image from RGB to a HSV image, and separated the individual component.

I was able to determine an average Hue value, but this is for the entire image. It would be more useful if I could get Hue values for each of the 9 individual circles. Is there a way to do this without me creating individual masks and applying them seperately?

I've attached part of the code below which I've seen in a previous answer, however is errors for the "isInsideCircle = distances <= Rad ;" line due to "Matrix dimensions must agree"

N.B. The images are screengrabs so please ignore the inconsistency of dimensions and positions. I added them just as a visual aid.

Variables

Rad = [110;110...] - This is the radius for all 9 circles

k=1:9 (Number of circles)

C(k,1) and C(k,2) are the X and Y coordinates for Circle k's centre. - This array has all coordinates for 9 circle centres.

XCo and YCo are the dimensions of my image (3712x5568)

for k = 1:numel(Rad)
    ROI2 = images.roi.Circle(gca); % create an ROI object
    ROI2.Radius = Rad(k); % specify its size and location
    ROI2.Center = C(k,:);
    for XCo = 1:3712
        for YCo = 1:5568
             distances(XCo,YCo) = sqrt( (XCo - C(k,1)).^2 + (YCo - C(k,2)).^2 );
             isInsideCircle = distances <= Rad ;
             counts = isInsideCircle ;
        end
    end
    
end

Answer 1

[Milan Bansal]

Hi Matthew,

I understand that you want to calculate the average hue value for each of the 9 circles in the image without creating the individual masks and applying them separately. Also

It is required to calculate the individual mask to get the average hue values for each circle, however "meshgrid" can be used to calculate the distances of pixels from the centre of each circle instead of iterating through all the X and Y coordinates in the entire image.

Please refer to the steps and pseudo code in the code snippet below:

% Extract the hue channel from the HSV Image
hueChannel = hsvImg(:,:,1);

% Initialize the array for average hue values
averageHues = zeros(1, 9);

% Create grid of coordinates
[X, Y] = meshgrid(1:3712, 1:5568);

% Loop through each circle to calculate the average hue
for k = 1:9
    % Calculate the squared distance from the center of the kth circle
    distSquared = (X - centers(k,1)).^2 + (Y - centers(k,2)).^2;
    
    % Find the indices of pixels that lie within the kth circle
    circleIndices = distSquared <= Rad(k)^2;
    
    % Calculate the average hue for the kth circle
    averageHues(k) = mean(hueChannel(circleIndices));
end

Please refer to the following documentation link to learn more about "meshgrid".

https://www.mathworks.com/help/matlab/ref/meshgrid.html

Hope this helps!

[Matthew]

Hi,

Thank you for your reply! Your code appears to function, however some of the values are very low/zero.

I managed to create masks for the individual tablets, and have made a very inefficient code to average the Hue. My only worry is this gives me different values compared to when using your code.

Tab1=BW2;

Circle1=images.roi.Circle(gca);

Circle1.Radius=110;

Circle1.Center=C(1,:);

TabMask1=createMask(Circle1, Tab1);

Tab1(TabMask1)=1;

Tab1(~TabMask1)=0;

Tab1RGB=bsxfun(@times, TabRGB1, cast(Tab1, 'like', TabRGB1));

Tab1HSV=rgb2hsv(Tab1RGB);

[Hue1, Sat1, Val1] =imsplit(Tab1HSV);

MeanHue(1,1)=mean(Hue1(:));

This set of code is repeated for each tablet. "MeanHue" is using my code and "averageHues" is using the code you supplied. Do you know which data set is accurate?

N.B. BW2 is the binary mask I supplied in my original post.