CODEWITHSUNDEEP
matlabmatrix
Alex L
Alex L

Reputation: 8925

Indexes and values of surrounding cells in 3d matrix

I'd like to return the indexes and values of the 8 cells surrounding a cell in a 3d matrix. 

mat = rand(5,5,5);

% Cell of interest
pos = [3 3 3]
pos_val = mat(pos(1), pos(2), pos(3))

% Surrounding cells
surrounding_pos = [pos(1)-1:pos(1)+1; pos(2)-1:pos(2)+1; pos(2)-1:pos(2)+1]
surrounding_val = mat(surrounding_pos(1,:), surrounding_pos(2,:), surrounding_pos(3,:))

This works fine for values in the centre of a matrix, but it breaks if pos is on the edge. (E.g. if pos was [3,4,5], surrounding_pos would include [3,4,6], which is out of bounds)

I could obviously remove surrounding_pos values <0 or >size(mat), but this doesn't seem like a terribly MATLABian method. Any ideas?

Upvotes: 5

Views: 926

Answers (3)

Kyle Wright
Kyle Wright

Reputation: 520

I found this post because I needed to grab the surrounding indices of a chosen point in a Matrix. It seems like the answers here are returning a matrix of the surrounding values, but the question is also interested in the surrounding indices. I was able to do this with "try/catch" statements, which I previously did not know existed in MATLAB. For a 2D Matrix, Z: 

%Select a node in the matrix
Current = Start_Node;

%Grab its x, y, values (these feel reversed for me...)
C_x = rem(Start_Node, length(Z));
if C_x ==0
    C_x =length(Z);
end
C_y = ceil(Start_Node/length(Z));
C_C = [C_x, C_y];

%Grab the node's surrounding nodes.
try
    TRY = Z(C_x - 1, C_y);
    Top = Current -1;
catch
    Top = Inf;
end
try
    TRY = Z(C_x + 1, C_y);
    Bottom = Current +1;
catch
    Bottom = Inf;
end
try
    TRY = Z(C_x, C_y + 1);
    Right = Current + length(Z);
catch
    Right = Inf;
end
try
    TRY = Z(C_x, C_y - 1);
    Left = Current - length(Z);
catch
    Left = Inf;
end

surround = [Top, Bottom, Left, Right];
m = numel(surround == inf);
k = 1;
%Eliminate infinites.

surround(surround==inf) =[];

I hope someone finds this information relevant.

Upvotes: 0

Colin T Bowers
Colin T Bowers

Reputation: 18560

Here is a tidied up version. Cheers.

mat = rand(5,5,5);
N = size(mat)
if length(N) < 3 || length(N) > 3; error('Input must be 3 dimensional'); end;
pos = [1 3 5]
surrounding_val = mat(max(pos(1)-1, 1):min(pos(1)+1, N(1)), max(pos(2)-1, 1):min(pos(2)+1, N(2)), max(pos(3)-1, 1):min(pos(3)+1, N(3)))

EDIT: Added an error trap.

Upvotes: 4

Gunther Struyf
Gunther Struyf

Reputation: 11168

Same solution as discussed here, but extended to multiple (any) dimensions:

mat = randi(10,5,5,5);
siz = size(mat );
N = numel(siz); % number of dimensions
M = 1; % surrounding region size

pos = [3 3 3];
pos_val = mat(pos(1), pos(2), pos(3));

surrounding_pos = cell(N,1);
for ii=1:N
    surrounding_pos{ii} = max(1,pos(ii)-M):min(siz(ii),pos(ii)+M);
end
surrounding_val2 = mat(surrounding_pos{:});

The important part is the last four lines, it avoids having to c/p the max, min thing for every dimension..

Or if you like short code, the loop changed to an arrayfun:

surrounding_pos = arrayfun(@(ii) max(1,pos(ii)-M):min(siz(ii),pos(ii)+M), 1:N,'uni',false);
surrounding_val2 = mat(surrounding_pos{:});

Upvotes: 5

Related Questions