The basic steps I'd use are:
- Perform a dilation on the shape to get a new area which is the shape plus its boundary
- Subtract the original shape from the dilated shape to leave just the boundary
- Use the boundary to index your data matrix, then take the minimum.
Dilation
What I want to do here is pass a 3x3 window over each cell and take the maximum value in that window:
[m, n] = size(A); % assuming A is your original shape matrix
APadded = zeros(m + 2, n + 2);
APadded(2:end-1, 2:end-1) = A; % pad A with zeroes on each side
ADilated = zeros(m + 2, n + 2); % this will hold the dilated shape.
for i = 1:m
for j = 1:n
mask = zeros(size(APadded));
mask(i:i+2, j:j+2) = 1; % this places a 3x3 square of 1's around (i, j)
ADilated(i + 1, j + 1) = max(APadded(mask));
end
end
Shape subtraction
This is basically a logical AND and a logical NOT to remove the intersection:
ABoundary = ADilated & (~APadded);
At this stage you may want to remove the border we added to do the dilation, since we don't need it any more.
ABoundary = ABoundary(2:end-1, 2:end-1);
Find the minimum data point along the boundary
We can use our logical boundary to index the original data into a vector, then just take the minimum of that vector.
dataMinimum = min(data(ABoundary));