Matrix indexing of a 3D matrix, one by one each 3D layer?

J = cat(3, magic(4), magic(4), magic(4));   % 3D matrix
maskIdx = J(:,:,1)>8;

%%I want to do something like below
J{:,:,1}(maskIdx) = ones(8,1);
J{:,:,2}(maskIdx) = 2*ones(8,1);
J{:,:,3}(maskIdx) = 3*ones(8,1);

Please focus on LHS only. How to apply same maskIdx on different layers. Note maskIdx created by the 1st layer which is 2D.

[Rik]

The only thing I can think of is using find and subs2ind to generalize this, but that feels very ineffective. So if nobody has a better idea and you can't implement this on your own, comment here and I'll post an answer with example code.

[JAI PRAKASH]

i understood ur idea.

I am looking if, it can be done in a miraculous way!! (I mean most efficient way)

Because if suppose J is a very big(10000x10000x3), then maskIdx will be 3 times. That's why I want to avoid the repetition.

Answer 1

[Rik]

Logicals are 8 times smaller than doubles, so memory concerns shouldn't be an issue. If you can work with a big J, you can easily afford to replicate the mask. The code below does its thing in 3.4 seconds for a 1000000x100x3 J.

1e5 done in 0.0 seconds
1e6 done in 0.0 seconds
1e7 done in 0.3 seconds
1e8 done in 3.4 seconds

    Name               Size                    Bytes  Class      Attributes

    J            1000000x100x3            2400000000  double               
    maskIdx      1000000x100x3             300000000  logical

code:

for k_base=5:8
    k=10^(k_base-2);
    clear J maskIdx n targetDim

      J=rand(k,100,3);
      maskIdx=J(:,:,1)>0.8;
      n=sum(maskIdx(:));
      A=ones(n,1);
      B=2*A;
      C=3*A;
      tic

      maskIdx(1,1,size(J,3))=false;%extend to 3D
      targetDim=1;
      J(circshift(maskIdx,targetDim-1,3))=A;
      targetDim=2;
      J(circshift(maskIdx,targetDim-1,3))=B;
      targetDim=3;
      J(circshift(maskIdx,targetDim-1,3))=C;

      fprintf('1e%d done in %.1f seconds\n',log10(numel(J)/3),toc)
  end