Create Small Neighborhood of image

I want to make small neighborhood of every pixel x. The small neighborhood has center x (pixel x be the center) in radius w.

Suppose w=5, so I have 5x5 small neighborhood with pixel x is the center of neighborhood.

How can I make it?

Thank you.

Answer 1

[Vinai Datta Thatiparthi]

Hello Yohanes,

val = imread(' ... '); % Your input matrix/image
[xCoOrd, yCoOrd] = [ ... ] % Co-Ordinates of the center pixel
nbh = 5; % Neighborhood Size

For a square neighborhood, the variable roi will hold the intensity values of neighboring pixels.

limit = floor(nbh/2);
roi = val(xCoOrd-limit:xCoOrd+limit, yCoOrd-limit:yCoOrd+limit);

Additionally, if you want the positions of the neighbors in subscripts,

rowLen = [xCoOrd - limit: xCoOrd + limit]'; % Range of Rows
colLen = [yCoOrd - limit: yCoOrd + limit]; % Range of Columns
rowIdx = meshgrid(rowLen, 1:nbh)'; % All the X coordinates of neighbors
colIdx = meshgrid(colLen, 1:nbh); % All the Y coordinates of neighbors

Hope this helps!

[Yohanes Setiawan]

Hello,

Thanks for your answer.

For this co-ordinates center pixel:

[xCoOrd, yCoOrd] = [ ...

What does "[ ..." means?

So if I want all pixels be the center, I just take "for loop" on it?

Also, I have ever tried some code like:

rowLen = [xCoOrd - limit: xCoOrd + limit]'; % Range of Rows

colLen = [yCoOrd - limit: yCoOrd + limit]; % Range of Columns

Then, MATLAB notices about array indices should be positive integers or logical values when I input:

xCoOrd= 1, yCoOrd=1 %this is the coordinate of pixel in val(1,1) as the center

How do I fix it?

[Vinai Datta Thatiparthi]

Hey Yohanes,

"What does "[ ..." means?"

The coordinates of the center pixel, of which you want to find the neighbors, can be placed inside the square brackets.

"So if I want all pixels be the center, I just take "for loop" on it?"

Yup, if you want to iteratively move through every pixel in your image, you might have to go ahead with a for loop.

"Then, MATLAB notices about array indices should be positive integers or logical values when I input:"

Yup, this is expected behaviour since indices go out of the positive range. You have to take care of these edge cases. Have a look at this MATLAB Answer: https://www.mathworks.com/matlabcentral/answers/86900-how-to-find-all-neighbours-of-an-element-in-n-dimensional-matrix

Hope this helps!

[Yohanes Setiawan]

Halo,

Thank you very much. Your answer is really help me.

For the edge cases, how if I use this:

a=xCoOrd-limit;
b=xCoOrd+limit;
c=yCoOrd-limit;
d=yCoOrd+limit;
[row col]= size(val);
if a<1
    a=1;
end
if b>row
    b=row;
end
if c<1
    c=1;
end
if d>col
    d=col;
end

Is that logical correct (especially in the neighborhood concept of image processing) for computing neighbor of a pixel? Or we need the padding as you give me the link for solution?

[Vinai Datta Thatiparthi]

Hey Yohanes,

Yes, this approach could work as well. Simple if-else statements checking for edge cases, and then assigning the neighbors (the ones that are beyond scope) NaN values.