getting statistics from within a mask within an image

We have an image that represents data that only makes sense when it is analyzed in numerical format.

Specifically, the data needs to be analyzed as a function of radius as shown below

Im interested specifically in hte max and min values within each of the defined areas with respect to the center point.

Ive been looking at a few examples online and it seems this should work when you pull the data from a mask.

However, the issue is that I seem to be getting values that are not realistic.

How to get pixel value inside a circle - MATLAB Answers - MATLAB Central (mathworks.com)

how to draw circle in an image? - MATLAB Answers - MATLAB Central (mathworks.com)

this is what I am doing

clear
img = double(imread('img121.jpg'));;  %no filtration  

img = -(0.0316*img) +8.3;   % we did this as we cant calibeate the film, we scan the same film over and over and it changes by 80pixels
img = imrotate(img, 90);
img = imgaussfilt(img ,1.5);
figure, imagesc(img )
axis image
height2 = 3.6;
caxis([0 height2])
colorbar
title(' ')
impixelinfo
%# make sure the image doesn't disappear if we plot something else
hold on
%https://www.mathworks.com/matlabcentral/answers/1931825-how-to-get-pixel-value-inside-a-circle
%below looks like what we want
%https://www.mathworks.com/matlabcentral/answers/1931825-how-to-get-pixel-value-inside-a-circle
%# define points (in matrix coordinates)
%3"
cpx = 2050;
cpy = 2020;
inchlist = [12,10.5,9,7.5,6,4.5];
%draw lines on heel axis
for n=1:size(inchlist,2)
    inch=  inchlist(n)/4;
    hcirc = drawcircle('Center',[2050,2020],'Radius',inch*590,'StripeColor','red');
    mask1 = hcirc.createMask;
    maxval = (max(img(mask1)));
    minval = (min(img(mask1)));
    uniformity = maxval/minval
    %  p1 = [cpy-100,cpx+inch*590];
end

uniformity = 6.8965

uniformity = 5.3893

uniformity = 5.3893

uniformity = 5.3893

uniformity = 3.7992

uniformity = 1.9168

Even after getting this max and min value, I will need to remove 10 to get rid of noise. Extra credit if you can point me to a soluton for that too.

thank you

[JM]

also interested in trying to understand how to debug this

how to i visualize (on the image) the area that are producing the improper values ?

was thinking...

[I,J] = max(max(img(mask1)))

I =

4.56390040017472

J =

1

[JM]

Anyone have any thoughts?

[JM]

all set

Answer 1

[Walter Roberson]

To find out where the maximum of the masked image is:

IdxImg = reshape(1:numel(img), size(img));
masked_image_idx = IdxImg(mask1);
[max_value, max_idx] = max(img(mask1));
index_of_max = masked_image_idx(max_idx);
[MaxRow, MaxColumn, MaxPage] = ind2sub(size(img), index_of_max);

This works by creating an indexing image the same size as the original image, but containing 1, 2, 3, ... all the way up to the maximum array element number. Then you extract the portion of the indexing image under the mask, to get a vector of the element numbers covered by the mask. Then take the max() of the masked image, getting out the (linear) index of the maximum element. Use the linear index of the maximum element to index the vector of indexing array elements selected by the max, getting out the linear index of the array element in the original image. Then convert the linear index back to subscripts if necessary.

[JM]

Walter,

I appreciate the help with this function to get the max and min values. I admit I dont totaly understand the masking.

Using the coordinates returned from the code you provided, Im still not able to see the pixel values being retuned as the max and min that make sense with the orignal image. There is something funny going on.

I have attached the orignal image that is being run with the script.

from scanning through the areas defined by the mask, I would expect each region to produce a max and min that produce uniformity (max/min) that is around 1 to 2. However, we are getting values of like 8 which is not possible. Yes, there are noise points in there but nothing to produce large values like this.

thank you again.