Efficient way to calculate distance matrix between cells

Hi,

So i have an image with size of 512x512 pixels

For each pixel i calculte an environment of 11x11 nighbours so now i have matrix of cells in size of 512x512 while each cell contain a matrix in size of 11x11.

i want to calculte in efficient way a distance matrix between the cell's matrix elements, i.e if i organzie the cells marix in colum stack (vector in size of 262144), lets call this vector A , (so A is vector of cells while each cell contain a 11x11 matrix) i would like to create a matrix D in size of 262144x262144 while D(i,j) is the euclidian distance between the matrices in A(i) and A(j),

So far i calculate it with for loops:

function [ D ] = DistanceMatFromPatches( patches )

    [patchesH,patchesW]=size(patches); 
     
    D=zeros(patchesH,patchesW);    
    patches_cs=patches(:);
    patchesSize=size(patches_cs);
    patchesNum=patchesSize(1);
    f = waitbar(0);
    for i=1:patchesNum-1
        first_cs=patches_cs{i}(:);
        for j=(i+1):(patchesNum)
            waitbar(j/patchesNum,f,sprintf('Calculate Distance matrix: pixels %d and %d from %d',i,j,patchesNum));
            %pause(0.1);
            sec_cs=patches_cs{j}(:);
            D(i,j)=norm(first_cs-sec_cs);
            D(j,i)=D(i,j);
        end 
    end
    close(f);
    
    
end

while patches is 512x512 matrix of cells.

This implementation is not very efficent .. anyone has other idea?

Answer 1

[Jan]

A 262144 x 262144 matrix of type double needs 550 GB of RAM. I assume that this will exhaust your machine.

You need a different approach, which can run on existing computers.

Blowing up the matrix to a cell matrix, which contains 11x11 submatrices does not add any new information to the data. Then this produced redundancy is a waste of ressources already.

[Orr Streicher]

Hi Jan and thanks for your response

Any seggestion how to do it?

In the end i would like to find the distance matrix as i described .. maybe there is an efficient way to do it?