benchmark_peak_local_max.py

import numpy as np

from scipy import ndimage as ndi
from skimage.feature import peak_local_max


class PeakLocalMaxSuite(object):

    def setup(self):
        mask = np.zeros([500, 500], dtype=bool)
        x, y = np.indices((500, 500))
        x_c = x // 20 * 20 + 10
        y_c = y // 20 * 20 + 10
        mask[(x - x_c)**2 + (y - y_c)**2 < 8**2] = True

        # create a mask, label each disk,
        self.labels, num_objs = ndi.label(mask)
        # create distance image for peak searching
        self.dist = ndi.distance_transform_edt(mask)

    def time_peak_local_max(self):
        local_max = peak_local_max(
            self.dist, labels=self.labels,
            min_distance=20, indices=False, exclude_border=False)