Fast Euclidean morphological operators using local distance transformation by propagation, and applications

We propose a new method to compute the morphological dilation of a binary image with a circular structuring element of any given size, on a discrete lattice. The algorithm is equivalent to applying a threshold on an exact Euclidean distance map, but computations are restricted to a minimum number of pixels. The complexity of this dilation algorithm is compared to the complexity of the commonly used approximation of circular structuring elements and found to have a similar cost, while providing better results.