Slide matrix on a larger matrix

hello,

I am to doing some masking.

I am trying to move a matrix(B) of undefine size could be any MxM on a larger matrix (A).

For example I have 10 by 10 matrix(B) fill with zeros and i want to move this matrix through out Matrix(A) 1000x1000.

This matrix(A) 1000X1000 is a binary matrix that have circles represented by 0.

I am stuck on how to do this nested for loops to make this movement, and how to find out one if the circles in matrix(A) overlaps with the matrix(B) if dont know what side this Matrix B will be after computation.

can't use function such as

[x,y]=bwlabel(I);

so far i have this set of steps:

I= imread('circles.JPG');

g = rgb2gray(I);

IB=imbinarize(g);

[ROWS,COLUMNS]=size(IB);

imshow(IB);

temp=0;

d=0;

k=0;

t=0;

A=ones(ROWS,COLUMNS)

for i=1:ROWS % find largest diameter

for j=1:COLUMNS

if IB(i,j)==0

temp=temp+1;

if temp >= d

d=temp;

end

elseif IB(i,j)==1

temp=0;

end

end

end

u=0;

v=0;

for i=1:ROWS

for j=1:COLUMNS

if IB(i,j)==0

B=zeros(d,1);

[r,c]=size(B);

for x=1:r

for y=1:c

if IB(I+u,j+v)==0 && B(r,c)==0

how to find the overlap??

end

end

end

end

end

end

Answer 1

[Mohammad Sami]

You should be using convolution for this instead of for loops.

you can use the conv2 function for 2d matrices.

https://www.mathworks.com/help/matlab/ref/conv2.html

[GIO COR]

thank you,

how i will be moving the smaller matrix to each section of the largest to do the convolution at each section of the largest image? or just do conv2(A,B) and all the circles with that match will disapear?

[Mohammad Sami]

you can see visually how convolution work at this link.

https://towardsdatascience.com/intuitively-understanding-convolutions-for-deep-learning-1f6f42faee1

[Mohammad Sami]

The smaller matrix is called a kernel (you can think of it as a filter) which you are applying on a bigger matrix (image) to identify a specific feature.