calculating the neighbour values

i have a matrix

A=[1 2 3

   4 5   6

   7  8  9  ]

please tell how to calculate the neighbour values of centre pixel

Answer 1

[Andrei Bobrov]

A = [    542         605         341         615         928
         471         550         839        1185         315
         945         798         809         421         625
         982         876         201         723         863
         231         932         147         277        1100]; %initial matrix

out1 = cell(numel(A),1);
q = false(size(A));
for ii = 1:numel(A)
    w = q;
    w(ii) = true;
    w1 = imdilate(w,[1 1 1;1 0 1;1 1 1]);
    out1{ii} = A(w1);
end

% other variant

B = padarray(A,[1 1],nan);
s = size(B);
addm = bsxfun(@plus,(0:2)',(0:2)*s(1));
ad = addm(:)';
C = reshape(1:prod(s),s);
C = C(1:end-2,1:end-2);
out2 = B(bsxfun(@plus,C(:),ad([1:4,6:9])));

% the centre pixel values is 809 and has index:

i1 = 13;
out1{i1}
ans =
         550
         798
         876
         839
         201
        1185
         421
         723
>> out2(i1,:)
ans =
         550         798         876         839         201        1185         421         723
>>

[kash]

A = randi(1234,5,5)

A =

         542         605         341         615         928
         471         550         839        1185         315
         945         798         809         421         625
         982         876         201         723         863
         231         932         147         277        1100

In this the centre pixel values is 809,i need 8 neighbours aroubd it

550 839 1185 798 876 201 723 421

[Andrei Bobrov]

Hi Kash! See my answer.

Answer 2

[Matt J]

>> A(5)=[]

A =

     1     4     7     2     8     3     6     9

[kash]

MAtt assume it may be large matrix 256x256,i just posted that for example

[Jan]

@kash: Please do not post a specific problem, when you want to solve a more general problem. This wastes the time of the ones, who want to help you.

Answer 3

[Matt J]

If you have the center pixel as a subscript index (m,n), then do

 B=A((-1:1)+m,(-1:1)+n);

 B(5)=[];

If you have the center pixel as a linear index, you can do as in the following example:

A =

    0.3380    0.3042    0.2004    0.3071
    0.8604    0.1598    0.5558    0.2813
    0.7002    0.6425    0.9985    0.5409
    0.2507    0.7799    0.2056    0.7081

 jumps=bsxfun(@plus,[0;1;2],[0 1 2]*size(A,2));
 jumps=jumps-jumps(5);
 jumps(5)=[];

Now, Suppose I want the neighbourhood of A(6)=0.1598. Then I would do

>> A(jumps+6)

ans =

    0.3380    0.8604    0.7002    0.3042    0.6425    0.2004    0.5558    0.9985

[kash]

i get error Index exceeds matrix dimensions.assume i have 257x257 matrix

[Matt J]

Note that the method only works for pixels with 8 neighbours. You haven't said how you want to handle pixels on the edge of the matrix.

[kash]

i need only 8 neighbours

[Image Analyst]

Let me repeat Matt's statement, which you just seemed to ignore: "You haven't said how you want to handle pixels on the edge of the matrix." In fact, if you just think about this, write down numbers near the edge of your matrix so you can visualize it, I think you'd be able to figure it out yourself.

[Matt J]

"i need only 8 neighbours"

You mean you think the error is triggered by a non-boundary pixel? Show us the larger code that you're using. Also, use DBSTOP to see the center pixel coordinates that are generating the error.