Extract structure within cell structure

Hello,

I need help extracting structures of difference sizes from an array. I want to to extract the structure into variables that go like this: stats1, stats2, stats3, stats4 and so on. I can do this line by line, but when I try to do this on a loop (because there will be 100 images) matlab won't run the code. I attached the images that I am working with

clear;
clc;
% Importing PNG images
for k = 1:3
  a{k} = sprintf('AB%d.png', k);
  AB{k} = imread(a{k});
  %spliting to RGB channels and image filtering
  [r{k},BW{k},b{k}] = imsplit(AB{k});
  BW{k}= imsharpen(BW{k});
  BW{k}=im2bw(BW{k},0.15);
  %imshow(BW{k})
  % getting images Stats
  stats{k}= regionprops(BW{k},"all");
end

%Below is what I need help creating a loop. This works fine, but I'd like
%to have a loop rather than writting the same line over and over
stats1 = stats(1,1);
stats1 = cell2mat(stats1);

stats2 = stats(1,2);
stats2 = cell2mat(stats2);

stats3 = stats(1,3);
stats3 = cell2mat(stats3);

%And so on... This code works for my needs, but I need a loop to create a new variable from stats1 to stats 3
%(there will be more images, like 100) 

I had this in mind, but it is also a cell with structures

for i=1
    stats1(i)= stats1(1,i);
end 

[Stephen23]

So far you have not given any reason why you need to force yourself into writing slow, complex, inefficient, buggy code which is hard to debug... when you already have a cell array (which you can trivially access using simple and efficient indexing):

https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

"I had this in mind, but it is also a cell with structures"

Note that your loop does nothing.

[Francisco]

Thanks for answering.

The following code works for my needs, I am trying to locate spheroids from images of nerve tissue

%MATLAB CODE FOR REGIONPROPS
clear;
clc;
A = imread('AB10.png');
figure,imshow(A)
[r,BW,b] = imsplit(A);
figure,imshow(BW);
BW = imsharpen(BW);
%figure,imshow(BW)
%BW = imreducehaze(BW);
%BW = histeq(BW);
%BW = imnoise(BW);

BW = im2bw(BW,0.15);
figure,imshow(BW)
stats = regionprops(BW,"all");

n=1;
for i = 1:length(stats)
    area(i)= stats(i).Area;
    perimeter(i) = stats(i).Perimeter;
    circularity(i)= stats(i).Circularity;
    centroid{i}= stats(i).Centroid;
    majoraxislength(i) = stats(i).MajorAxisLength;
    minoraxislength(i) = stats(i).MinorAxisLength;
    if circularity(i)~=Inf && circularity(i)>0.47 && area(i)>150
        large_areas(n) = area(i);
        large_perimeter(n) = perimeter(i);
        large_circularity(n) = circularity(i);
        large_centroid(n) = centroid(i);
        large_majoraxislength(n) = majoraxislength(i);
        large_minoraxislength(n) = minoraxislength(i);
        n= n+1;
    end 
end

centers = large_centroid';
centers = cell2mat(centers);
large_minoraxislength = large_minoraxislength';
large_majoraxislength = large_majoraxislength';

large_Stats = table(large_areas',large_perimeter', large_centroid', large_circularity', large_majoraxislength, large_minoraxislength);
large_Stats.Properties.VariableNames(1:4) = {'Area' 'Perimeter' 'Centroid' 'Circularity'};


diameters = mean([large_majoraxislength large_minoraxislength],2);
radii = diameters/2;
hold on
viscircles(centers,radii);
hold off

This code is for only one image. What I am tyring to do, is for my code to process multiple images at once

Answer 1

[Kartikay Sapra]

To achieve the given task, create a custom function (say 'analyse') which returns a struct (say 'out') and fields of 'out’ can be your desired 'stats'.

After completing the above stated workflow, create an empty cell array with size corresponding to number of images (say 'stats')

This workflow will look like:

n = 100; %assuming number of images is 100
stats = cell(n, 1);
for k = 1:n
    stats{k} = analyse(k);
end

And your function may look like:

function out = analyse(k)
%MATLAB CODE FOR REGIONPROPS
    filename = "AB" + k + ".png";
    A = imread(filename);
    figure,imshow(A)
    [r,BW,b] = imsplit(A);
    figure,imshow(BW);
    BW = imsharpen(BW);

    BW = im2bw(BW,0.15);
    figure,imshow(BW)
    stats = regionprops(BW,"all");

    n=1;
    for i = 1:length(stats)
        out.area(i)= stats(i).Area;
        out.perimeter(i) = stats(i).Perimeter;
        out.circularity(i)= stats(i).Circularity;
        out.centroid{i}= stats(i).Centroid;
        out.majoraxislength(i) = stats(i).MajorAxisLength;
        out.minoraxislength(i) = stats(i).MinorAxisLength;
        if out.circularity(i)~=Inf && out.circularity(i)>0.47 && out.area(i)>150
            out.large_areas(n) = out.area(i);
            out.large_perimeter(n) = out.perimeter(i);
            out.large_circularity(n) = out.circularity(i);
            out.large_centroid(n) = out.centroid(i);
            out.large_majoraxislength(n) = out.majoraxislength(i);
            out.large_minoraxislength(n) = out.minoraxislength(i);
            n= n+1;
        end 
    end

    out.centers = out.large_centroid';
    out.centers = cell2mat(out.centers);
    out.large_minoraxislength = out.large_minoraxislength';
    out.large_majoraxislength = out.large_majoraxislength';

    out.large_Stats = table(out.large_areas',out.large_perimeter', out.large_centroid', out.large_circularity', out.large_majoraxislength, out.large_minoraxislength);
    out.large_Stats.Properties.VariableNames(1:4) = {'Area' 'Perimeter' 'Centroid' 'Circularity'};


    out.diameters = mean([out.large_majoraxislength out.large_minoraxislength],2);
    out.radii = out.diameters/2;
    hold on
    viscircles(out.centers,out.radii);
    hold off
end

Now for each image ‘ABk.png’ you extract 'stats' as:

stats{k}

where k is the index of the image.

Furthermore, the fields can be viewed by using the fieldnames:

%example of 'area'
stats{k}.area
%and so on