matlab求图像的局部峰值点集

经过辛苦寻找，目前发现三种方法来做（其中alpha控制强度）：
方法一：
I = imread('lena-gray.bmp');
%BW = nlfilter(I, [3 3], @(x) all(x(5) > x([1:4 6:9])+ alpha ) ); %方法1
imshow(BW);
This  "find strict max"  function would simply check if the center of the neighborhood is strictly greater than all the other elements in that neighborhood, which is always 3x3 for this purpose.

方法二：
I = imread('lena-gray.bmp');
%BW = imregionalmax(I);%方法2
imshow(BW);

The variable BW will be a logical matrix the same size as y with ones indicating the local maxima and zeroes otherwise.

NOTE: As you point out, IMREGIONALMAX will find maxima that are greater than or equal to their neighbors. If you want to exclude neighboring maxima with the same value (i.e. find maxima that are single pixels), you could use the BWCONNCOMP function. The following should remove points in BW that have any neighbors, leaving only single pixels:

CC = bwconncomp(BW);
for i = 1:CC.NumObjects,
index = CC.PixelIdxList{i};
if (numel(index) > 1),
BW(index) = false;
end
end


方法三：
I = imread('lena-gray.bmp');
BW = I > imdilate(I, [1 1 1; 1 0 1; 1 1 1]) +alpha ; %方法3
imshow(BW);

@Nathan: IMDILATE operates on each pixel of the grayscale image. The center of the 3-by-3 matrix is positioned at each pixel, and the pixel value is replaced by the maximum value found at the neighboring pixels where there is a value of 1 in the 3-by-3 matrix. The call to IMDILATE therefore returns a new matrix where each point is replaced by the maximum value of its 8 neighbors (zero padded at the edges as needed), and the points where the original matrix is larger indicates a local maxima.

imdilate goes over each pixel and computes the max of the neighboring pixels centered around it and specified by the mask given (notice the zero in the middle to exclude the pixel itself). Then we compare the resulting image with the original to check whether each pixel is strictly greater than the max of its neighborhood. Make sure to read the documentation page on morphological operations:

注：转自http://hi.baidu.com/lewutian

转载于:https://www.cnblogs.com/wzhwpersonalzone/p/3750434.html