matlab - how to find local maxima in image -

the question feature detection concept. i'm stuck after finding corner of image , want know how finding feature point within computed corners.

suppose have grayscale image have data this

a = [ 1 1 1 1 1 1 1 1;       1 3 3 3 1 1 4 1;       1 3 5 3 1 4 4 4;       1 3 3 3 1 4 4 4;        1 1 1 1 1 4 6 4;       1 1 1 1 1 4 4 4] 

if use

b = imregionalmax(a); 

the result this

b = [ 0 0 0 0 0 0 0 0;       0 1 1 1 0 0 1 0;       0 1 1 1 0 1 1 1;       0 1 1 1 0 1 1 1;       0 0 0 0 0 1 1 1;       0 0 0 0 0 1 1 1] 

the question how pick highest peak inside max local region (in sample how did chose 5 3 , 6 4)?

my idea using b detect each region , use imregionalmax() again i'm not @ coding , need advice or other ideas.

there couple of other easy ways implement 2d peak finder: ordfilt2 or imdilate.

ordfilt2

the direct method use ordfilt2, sorts values in local neighborhoods , picks n-th value. (the mathworks example demonstrates how implemented max filter.) can implement 3x3 peak finder ordfilt2 by, (1) using 3x3 domain does not include center pixel, (2) selecting largest (8th) value , (3) comparing center value:

>> mask = ones(3); mask(5) = 0 % 3x3 max mask =      1     1     1      1     0     1      1     1     1 

there 8 values considered in mask, 8-th value max. filter output:

>> b = ordfilt2(a,8,mask) b =      3     3     3     3     3     4     4     4      3     5     5     5     4     4     4     4      3     5     3     5     4     4     4     4      3     5     5     5     4     6     6     6      3     3     3     3     4     6     4     6      1     1     1     1     4     6     6     6 

the trick compare a, center value of each neighborhood:

>> peaks = > b peaks =      0     0     0     0     0     0     0     0      0     0     0     0     0     0     0     0      0     0     1     0     0     0     0     0      0     0     0     0     0     0     0     0      0     0     0     0     0     0     1     0      0     0     0     0     0     0     0     0 

imdilate

image dilation done on binary images, grayscale image dilation max filter (see definitions section of imdilate docs). same trick used ordfilt2 applies here: define neighborhood not include center neighborhood pixel, apply filter , compare unfiltered image:

b = imdilate(a, mask); peaks = > b; 

note: these methods find single pixel peak. if neighbors have same value, not peak.