Image processing code debugging

Hi, I am doing image processing and plans to extract some information from a binary image, however I have encountered some bugs that I can't solve.

Here is the error:

Error using double

Conversion to double from struct is not possible.

Error in ^ (line 44) (the thing is that my code only has 43 lines I dont kno how this error show up)

X = double(X);

Error in features (line 16)

C = P^2/(4 * pi * A);

Here is the code:

% To compute the features;
% input is the binary thresholded image 
% outputs are the feature values
function [P, A, C, xbar, ybar, phione] = features(Iin)
    %perimeter
    P = regionprops(Iin, 'Perimeter'); %regionprops computes the perimeter 
                                       %by calculating the distance between 
                                       %each adjoining pair of pixels around 
                                       %the border of the binary image.
    %area
    A = sum(sum(Iin == 255)); %summing the number of pixel that are white
    
    %compactness
    C = P^2/(4 * pi * A);
    
    %centroid
    [rows cols] = size(I);
    m = zeros(rows, cols);
    for i = 0:1 
        for j = 0:1 
            for x=1:rows 
                for y=1:cols 
                    m(i+1, j+1) = m(i+1, j+1) + (x^i * y^j * I(x,y)); 
                end
            end
        end
    end
    xbar = m(2,1) / m(1,1); %m10/m00
    ybar = m(1,2) / m(1,1); %m01/m00

    %invariant moment
    n=[ 0 0 0 0; 0 0 0 0; 0 0 0 0; 0 0 0 0]; 
    for i = 0:3 
        for j = 0:3 
            n(i+1,j+1) = u(i+1,j+1)/(u(1,1)^(1+(i+j)/2)); 
        end
    end
    phione = n(3,1) + n(1,3); %n(2,0) + n(0,2)
end

[Miles Hao Peng Su]

Aye thanks! I checked it up and replaced with the following line,

P = sum(sum(bwperim(Iin)==1)); 

Now it works!

Thanks!

[Image Analyst]

No. Don't do it that way! Didn't you see my answer below???

Answer 1

[KSSV]

This line:

P = regionprops(Iin, 'Perimeter');

Check the class of P

class(P)

It seems it is a structure and you can use it in the line:

 C = P^2/(4 * pi * A);

So check P and extract the field you want and substitue it in the line.

[Miles Hao Peng Su]

Aye thanks! I checked it up and replaced with the following line:

P = sum(sum(bwperim(Iin)==1)); 

Now it works!

Thanks!

Answer 2

[Image Analyst]

Scrap all that. Your Compactness is called "Circularity" and you can ask regionprops for the centroids, etc.:

function [P, A, C, xbar, ybar, phione] = features(Iin)
props = regionprops(Iin, 'Circularity', 'Centroid', 'Area', 'Perimeter');
A = [props.Area]; % List of all blob areas.
P = [props.Perimeter]; % List of all blob perimeters.
C = [props.Circularity]; % Circularities = (4*Area*pi)/(Perimeter^2)
xy = vertcat(props.Centroid); % List of all (x,y) centroid coordinates.  Each blob's coordinates is in a row.
xbar = xy(:, 1); % x coordinate of the blob centroids.
ybar = xy(:, 2); % y coordinate of the blob centroids.

[Miles Hao Peng Su]

Thanks for the help! I am currently learning image processing and can't use most of the off the shelf codes >< I will save this page when I need to use them later! Thanks!

[Image Analyst]

But I didn't use any more "off the shelf" functions than you. In fact, I used fewer. You used regionprops like me but you also used - unnecessarily I might add -- bwperim. I don't know what functions you are allowed to use or not use, but you used regionprops so I assume that's allowed and that's all I used. I assume non-toolbox functions like vertcat and sum are allowed.

There is a lot that can be improved with your codes. For example there might be multiple blobs in general so you can't do what you did. You should do

P = regionprops(Iin, 'Circularity', 'Centroid', 'Area', 'Perimeter');
A = [P.Area]; % List of all blob areas.
allPerims = [P.Perimeter]; % List of all blob perimeters.
C = allPerims.^2 ./ (4 * pi * A); % Doing it yourself instead of useing regionprops.

You need to use use the ".Perimeter" field and use square brackets around it do collect perimeters from (possibly) multiple blobs into one list (vector). That's because your P is a structure, even if it's just one blob you are looking at.

Please see my Image Processing Tutorial in my File Exchange:

Image Analyst's File Exchange

I think you could learn much from it.

If I've helped you consider clicking the "Vote" icon for my answer. I think it's a better approach than what you suggested or accepted.