Getting values from 4d matrix according (using) 2d matrix with logical values (0 or 1) - does not work as wanted

Hello!

Let say I have a matrix A of dimensions (50*100*248*20) with data and I have another one B (50*100) with logical values (lets image I have 100 of "1") and I want to get values from my 4D matrix according to B. So I am doing

C = A(B);

But then I am getting my values only from A (:,:,1,1) - first page, but I am expecting to get values from all the pages (along all the 3rd and 4 th dimensions). How can I do this?

So I need at the output a matrix with dimensions (100, 1, 248, 20) - which give me 100 values from each page of 248*20

If I am not clear, let me know!, Hoping for you help! Thanks!

[madhan ravi]

A(B,:,:) % totally a hunch

[Walter Roberson]

temp = reshape(A, size(A,1)*size(A,2), 1, size(A,3), size(A,4));
C = temp(B,:,:,:);

[Walter Roberson]

madhan, unfortunately your hunch will not work. B will effectively be reshaped, as if you had asked for A(B(:),:,:) . So the 50*100 B will act like a 5000 long logical index in the first dimension of A. That will fail if B contains any "true" value past the 50th entry, as it would be an index out of range.

The situation would have been different if you were using B as the last index, something like A(:,:,B) . When you pass in fewer index locations than there are dimensions, then MATLAB acts as-if the remaining dimensions were reshaped, as if you were accessing

reshape(A, size(A,1), size(A,2), size(A,3)*size(A,4)) indexed at (:,:,B(:))

Another way of saying that is that MATLAB uses linear index calculations for the last index that you provide, but not for the earlier ones.

[Yurii Iotov]

Walter Roberson, Thanks for the solution!

[madhan ravi]

Indeed sir Walter, thank you for the explanation just noticed the comment.

[Linus Olofsson]

Could someone please give an example of the explanation Walter gave above? I do not really understand how this is a solution of the problem, nor do I understand the code snippet provided by the earlier comment by Walter.

Could the solution please be explained with my very similar non-functional problem with a 3 channel color image and a 2 dimensional logical array (mask) that I would like to use to color code speficic parts of my image, i.e.:

image = zeros(150,5370,3);
whos mask
  Name              Size               Bytes  Class      Attributes
  mask            150x5370            805500  logical
image(mask,2) = 255; % Change the masked part of the image to be green.

Answer 1

[Yurii Iotov]

Thank you guys for being active and for your help!

What I would like to do also and kindly ask you to help is how to ''cut'' and keep the values in a matrix ''form''? I mean to get a values which corresponds to logical 1, to put them in a matrix and at this matrix will stay some neighboring ''non-desired values'' (which corresponds to 0, since we are cutting as a 'rectangular') but to replace this values to NaN. Primitive illustration:

B = [0 0 0 0 0
     0 1 1 1 0  =>  B1 = []
     1 1 1 0 0
     0 0 1 0 0]
 

.

=>  B1 = 0 1 1 1 =>  C = A(B1)
         1 1 1 0 
         0 0 1 0 

.

A = NaN Val Val Val
    Val Val Val NaN 
    NaN NaN Val NaN

Finally to plot A using pcolor or contour and to get smth like this (white zone represents these logical zeros, while plotted area logical ones and corresponding values)

Sorry if not clear...Thanks

[Walter Roberson]

If there are rows or columns that are all zero, then do you want them to be removed if they are in the middle, or do you only want to remove rows and columns of 0s that are at the edges ?

used_cols = find(any(B,1));
used_cols = used_cols(1):used_cols(end);
used_rows = find(any(B,2));
used_rows = used_rows(1):used_rows(end);
B1 = B(used_rows, used_cols);
A1 = A(use_rows, used_cols);
A1(~B1) = nan;
image(A1);
colormap(Appropriate_Colormap_Goes_Here)

You do not appear to be using any of the standard color maps.

[Yurii Iotov]

Zeros can only remains at the edges... logical 1s represent discretized (according to grid) shape of poligons (for example imagine hexagone or other complex shape)....Then from my 4d matrices with data I cut new ones with data which are inside this polygons (logical ones) and then I need to plot (colormap...etc) this data, but the ammount of zeros which was cut also I want to replace as NaN

Thanks

[Walter Roberson]

I think you mean that Zeros are only to be removed at the edges.

The code I posted here finds the bounding box that contains all of the 1's. It could easily be extended to more dimensions.

Answer 2

[Darren Wethington]

See if this is what you're looking for:

A = rand(10,20,30,40);
B = round(rand(10,20));

four_dim_logical = zeros(size(A));
for i=1:length(four_dim_logical(1,1,:,1))
    for j=1:length(four_dim_logical(1,1,1,:))
        four_dim_logical(:,:,i,j) = B;
    end
end
C = A(logical(four_dim_logical));

dim1 = sum(B(:));
dim2 = 1;
dim3 = size(A,3);
dim4 = size(A,4);

C = reshape(C,dim1,dim2,dim3,dim4);

[Darren Wethington]

Just saw solutions in the comments. This solution will be MUCH more computationally expensive if either of those work.

[Yurii Iotov]

Thanks!