Locally adaptable mathematical morphology

We investigate how common binary mathematical morphology operators can be adapted so that the size of the structuring element (SE) can vary across the image. We show that when the SE are balls of a metric, locally adaptable erosion and dilation can be efficiently implemented as a variant of distance transformation algorithms. Opening and closing are obtained by a local threshold of a distance transformation, followed by the adaptable dilation.