How to create a multidimensional matrix from a cell array where the cell array sizes are not he same?

I have a cell array C

C{1,1} = {[1 2 3]}
C{1,2} = {[12 21 3]}
C{1,3} = {[12 21 3], [43 45 66]}
C{1,4} = {[2 5 1], [1 65 3], [32 5 1], [2 5 1]}

C{2,1} = {[1 2 3], [11 6 43], [1 5 1], [4 56 1]}
C{2,2} = {[12 21 3]}
C{2,3} = {[43 45 66]}
C{2,4} = {[1 65 3], [32 5 1], [2 5 1]}

C{3,1} = {[4 56 1]}
C{3,2} = {[12 21 3], [23 5 2], [2 4 2]}
C{3,3} = {[1 4 2], [43 45 66]}
C{3,4} = {[37 45 6]}

As can be seen, the cells are of different dimension. Each element of a cell has values like [x y z]

Now I want to convert this to a multidimensional matrix of dimension 3*4*4(in the above example) with the missing entries in the matrix to be [0 0 0].

C_New(1,1,:) = [1 2 3; 0 0 0; 0 0 0; 0 0 0]  
C_New(1,2,:) = [12 21 3; 0 0 0; 0 0 0; 0 0 0]  
C_New(1,3,:) = [12 21 3; 43 45 66; 0 0 0; 0 0 0]  
C_New(1,4,:) = [2 5 1; 1 65 3; 32 5 1; 2 5 1]
.
.
.
.
C_New(3,1,:) = [4 56 1; 0 0 0; 0 0 0; 0 0 0]  
C_New(3,2,:) = [12 21 3; 23 5 2; 2 4 2; 0 0 0]  
C_New(3,3,:) = [1 4 2; 43 45 66; 0 0 0; 0 0 0]  
C_New(3,4,:) = [37 45 6; 0 0 0; 0 0 0; 0 0 0]

what is the best way to do this? I want to do this because having a matrix makes it easy for me to vectorize my operations and gain speed. My actual cell arrays are much bigger (900*3600).

Any help is appreciated! :)

[Azzi Abdelmalek]

your example is not clear. Post a usable example

[Andrei Bobrov]

? please example:

C{1,3} = {[12 21 3], [43 45 66]};

or

C{1,3} = [12 21 3; 43 45 66] ?

[Amulya NV]

It is the fist one. Question edited

[Andrei Bobrov]

? maybe so:

C_New(:,:,1,1) = [1 2 3; 0 0 0; 0 0 0; 0 0 0];
.
.
C_New(:,:,3,3) = [1 4 2; 43 45 66; 0 0 0; 0 0 0]

[Amulya NV]

Yes.. You are right!

[Andrei Bobrov]

Please see my answer.

Answer 1

[Andrei Bobrov]

EDIT

n = cellfun(@numel,C);
nmx = max(n(:));
K = cellfun(@(x,y)[cat(1,x{:});zeros(nmx-y,3)],C,num2cell(n),'un',0);
CC = cat(3,K{:});
C_New = reshape(CC,[nmx,3,size(C)]);

[Amulya NV]

Works like magic! Thanks a lot :)

Answer 2

[Walter Roberson]

Zpad3 = @(V) [V(:);zeros(3-length(V),1)];

Result = cell2mat(permute(cellfun(Zpad3, YourCell, 'Uniform', 0), [3 1 2]));

You might be able to do a little better on efficiency by using cat() to put portions together. You would look at the code for cell2mat and see if you could tweak it slightly so that you did not have to do the permute() while putting the slices together.

The permute() is there to reshape the array to be length 1 in the first dimension, ready to be expanded to length 3 by the length 3 column vector (column vector is occupied in the first dimension.)