The Image Processing Toolbox has several functions that, when used together, can identify user-defined patterns in masked images. While there are many different ways of accomplishing this in MATLAB. The following example is a simple method that can be applied in many cases:
im1 = imread('image.png');
% binary matrix containing the outline of the desired feature etched out by 1's
mask = [...];
se = strel('arbitrary', mask);
% apply a Sobel filter to the image which highlights the edges
im_f = edge(im1, 'sobel');
% filter the image using the stencil mask, if the stencil is good, it will
% only retain mask-shaped features in the Solbel-filtered image
im_er = imerode(im_f, se);
% enhance the features from the eroded image
im_di = imdilate(im_er, se);
% get the "center of mass" of each individual object in the enhanced,
% filtered image -- this works best if the original image was converted to
% black and white
rp = regionprops(im_di,'centroid');
After executing this example, the "rp" variable will contain the centroid locations of each instance of the feature described by "mask". Typically, "mask" is determined experimentally to achieve the desired results.
