Scan exactly 1 pixel around a location and determine max value

If I have a list of coordinates xf,yf identifying a bright object as in the image (red dots only)

[xf,yf] =
       4    10
       7     5
       7    15
      10    10
      13     5
      13    15
      16    10

It doesn't always correspond to the max intensity of the object. How do I go about scanning round the red dot by 1 pixel and then return the max value. I would need to keep the same order as the original xf,yf data.

Answer 1

[Image Analyst]

Call imdilate() which is a function that identifies the local max in a window:

maxImage = imdilate(grayImage, true(3));

To find the max at any particular (row, column) location from your list of red spots (in 1-D arrays called "xf" and "yf"), do

for k = 1 : length(xf)
    thisRow = yf(k);
    thisCol = xf(k);
    fprintf('The Max in the 3x3 neighborhood of row=%d, col=%d is %d\n',...
        thisRow , thisCol, maxImage(thisRow, thisCol));
end

[Jason]

Thanks IA, very elegant!

Answer 2

[Guillaume]

There are many ways you could do this (e.g using loops). Here is a possible solution using bsxfun:

imagewidth = 100;   %for e.g. required for bounds checks
imageheight = 200;  %for e.g. required for bounds checks
img = randi(256, imageheight, imagewidth) - 1; %for e.g.
xy = [
     4    10
     7     5
     7    15
    10    10
    13     5
    13    15
    16    10]; %xy = [xf yf]
%create cartesian product of offsets [-1 0 1] in both x and y direction
[dx, dy] = ndgrid(-1:1); 
%create 3d matrix of pixel coordinates by using offsets dx and dy
pixels = bsxfun(@plus, xy, reshape([dx, dy], 1, 2, []));
%each (:, :, k) matrix correspond to the xy matrix shifted by [dx(k), dy(k)]
%make sure that pixels are within image bounds:
pixels = max(pixels, 1);
pixels = min(pixels, repmat([imagewidth, imageheight], size(xy, 1), 1, numel(dx)));
%get maximum along third dimension of image pixels indexed by pixels:
maxpix = max(img(pixels(:, [2 1], :)), [], 3)

[Jason]

Thankyou