Local binary pattern varient

Hey everyone, i'm trying to write a code for a method of LBP which consists to calculate the diference between 2 neighboor pixell of each pixel sourrounding the center on a 3x3 window. and then assigned to each neighboring pixel a binary value 1 if the both neighboor pixels values are positive or negative , and 0 if one of them is positive and the other is negative. here's the code :

A = imread('inputImage.png');
grayImage = rgb2gray(A);
[rows, columns, numberOfColorBands] = size(grayImage);

localBinaryPatternImage1 = zeros(size(grayImage));

for row = 2:rows - 1
    for col = 2:columns - 1
        Ic = grayImage(row, col);
        I(6)=grayImage(row-1, col-1);
        I(7)=grayImage(row-1, col);
        I(8)=grayImage(row-1, col+1);
        I(1)=grayImage(row, col+1);
        I(2)=grayImage(row+1, col+1);
        I(3)=grayImage(row+1, col);
        I(4)=grayImage(row+1, col-1);
        I(5)=grayImage(row, col-1);
        
        n = 1:8; % n is the number of neighboor pixels
        if n == 1 
            k1(n) =I(8)-I(n); % i'm trying to calculate the difference between the 2 adjacent pixels of the neighboor pixel 
            k2(n) = I(n+1)- I(n);
            
        elseif n == 8
            k1(n) = I(n-1)- I(n);
            k2(n) = I1 - I(n);
            
        else
            n = 2:7 ;
            k1(n) = I(n-1)- I(n);
            k2(n) = I(n+1)- I(n);
        end 
        
       if k1(n)>= 0 & k2(n)>=0 % here i'm sitting the conditions to get a final binary code for each pixel 
           I(n)= 1;
       elseif k1(n)<0 & k2(n)<0
           I(n)= 1;
       elseif k1(n)>= 0 & k2(n)<0
           I(n)= 0;
       elseif k1(n)< 0 & k2(n)>= 0 
           I(n) = 0;
       end
         % i want to calulate the LBPD, multiplying the weights by the binary code gotten 
        LBPD = (I(1) + I(2) *2^7 + ...
            I(3) * 2^(6) + I(4) * 2^5 + ...
            I(5) * 2^4 + I(6) * 2^3 + ...
            I(7) * 2^2 + I(8)*2);
        
       localBinaryPatternD(row, col) = LBPD;        
                
    end
end

is it the right way to do it ? how can i fix it ? can anyone please help

Answer 1

[Image Analyst]

I didn't delve into your code but if you want to see how I did it, see my attached demo.

[nissrine Neyy]

Thank you @Image Analyst for your prompt response, i cheked your code it is clear and siplified way to obtain the LBP, although for my case i have more calculating to do since i'm calculating the differences between adjacent pixels of each of the 8 pixel and then assigning binary values depending on whether they're positive or negative. and it's where i pretty much find difficulties to get it done

[Image Analyst]

To find differences you can use conv2(). Like to find the average difference at each pixel you can do

kernel = [-1, -1, -1; -1, 8, -1; -1, -1, -1] / 8;
outputImage = conv2(double(inputGrayImage), kernel, 'same');

The output will be an image where the pixel value is the average difference between a pixel and its 8 closest neighbors.

[nissrine Neyy]

Maybe i misexplained the issue, the differences i'm trying to compute are the diferences between each of the 8 pixels and their 2 adjacents (not a pixel and it's 8 closest neighbors) here's an image to show the idea

[Image Analyst]

You could do that with a set of convolutions, or manually within a nested for loop. See attached demo and adapt the inner part of the loop to do this subtraction thing you want to do.