Follow:

https://www.mathworks.com/matlabcentral/answers/278353-2d-convolution-algorithm-error-vs-conv2-function

In Python:

https://stackoverflow.com/questions/3731093/is-there-a-python-equivalent-of-matlabs-conv2-function

Follow:

https://www.mathworks.com/matlabcentral/answers/278353-2d-convolution-algorithm-error-vs-conv2-function

In Python:

https://stackoverflow.com/questions/3731093/is-there-a-python-equivalent-of-matlabs-conv2-function

see link

https://www.mathworks.com/matlabcentral/answers/108868-edge-detection-using-sobel-operator

https://angeljohnsy.blogspot.com/2011/12/sobel-edge-detection.html

you may use this primitive and simplified code to achieve 2D convolution between an image and a given template.

[irows, icols] = size(image);

[trows, tcols] = size(template);

temporarry(1:irows,1:icols) = 0;

trhalf = floor (trows/2);

tchalf = floor (tcols/2);

for y = trhalf+1 : irows-trhalf

for x = tchalf+1 : icols-tchalf

sum = 0;

for jwin = 1:trows

for iwin = 1:tcols

sum = sum +image(y+jwin-trhalf-1, x+iwin-tchalf-1)*template(jwin, iwin);

end

end

temporarry(y, x) = sum;

end

end

conv2 is doing the filtering operation. That is nothing but the mask is to placed over a region of image and corresponding intensity value and mask coefficient value get multiplied and summed up which give a new value. And the center pixel will get replaced by summed up value. Now get the mask shift one location right and perform same operation. Finally you will get replaced all the pixel value. You need two FOR loop to write the code and keep in mind the corner pixel will not fit for this operation because the mask can not be placed over there.