3D-Image Processing find() several values performance

Basically I am trying to optimize this minimal example:

I am aware of histc() or histcounts(), however I require a linear index.

in contrast to this example the count of indices{} varies.

% Original size is 600x600x600
s=100;
% A is a binned greyscale image
A=repmat(magic(s),[1 1 s]);
% b is basically the bins
b=unique(A);

indices=cell(length(b),1);
tic
parfor i=1:length(b)  
    indices{i}=find(A==b(i));    
end
toc

Performance is really poor on 16 cores. What am I missing?

Also tried replacing find with the mex function from fileexchange

https://www.mathworks.com/matlabcentral/fileexchange/49162-fastfindk

Happy for all suggestings, I have a full license.

Cheers,

matt

Answer 1

[Kritika Bansal]

Hi,

You can try creating a parallel.pool.Constant and see if that helps you. You can find more information on the link below:

https://www.mathworks.com/help/parallel-computing/parallel.pool.constant.html;jsessionid=ad12ff1a6d5ad8f39a79fdd2b4ca