How can I find the top 5% values' index in a 3D matrix

I want to find the top 5% values' index in a 3D matrix

Answer 1

[Thorsten]

 X = rand(10, 10, 3); % sample data
 [~, idx] = sort(X(:));

Linear indices:

 idxtop5perc = idx(end-5*numel(X)/100+1:end)

Subscripts:

 [i, j, k] = ind2sub(size(X), idxtop5perc);

[Yun Shen]

some values in x maybe same, this can solve the problem?

[Thorsten]

If you want the indices of the top 5% values where some values can be the same, you can use

 u = unique(X(:));
 v = u(floor(end-5*numel(u)/100));
 ind = find(X >= v);

Answer 2

[Guillaume]

demodata = rand(100, 50, 20); %a 3d matrix;
[values, idx] = sort(demodata, 'descend');
%depending on what is meant by top 5%:
%5% of the population
idxpop = idx(1:ceil(end/20)) 
%5% of the maximum value
idxval  = idx(values >= values(1)*.95)

sort your data and filter it any way you want.