'''
Module to find local maxima in an image.
'''
import numpy as np
import scipy.ndimage.filters
import matplotlib.pyplot as plt
[docs]def local_max(img, r, thresh):
'''Find local maxima from comparing dilated and eroded images.
Calculates images with maximum and minimum within given radius. If the difference is larger than the threshold, the original pixel position with max value is detected as local maximum.
Parameters:
img (np.ndarray): Input image.
r (int): Radius for locality.
thresh (int/float): Intensity difference threshold.
Returns:
(np.ndarray): Array of points.
'''
try:
r = int(r)
thresh = int(thresh)
assert(isinstance(img, np.ndarray))
except:
raise TypeError('Bad input!')
# prepare circular kernel
y,x = np.ogrid[-r:r+1, -r:r+1]
kernel = x**2 + y**2 <= r**2
# calculate max and min images
img_dil = scipy.ndimage.filters.maximum_filter(img, footprint=kernel)
img_ero = scipy.ndimage.filters.minimum_filter(img, footprint=kernel)
# get selection of local maxima
sel = (img==img_dil)*(img-img_ero > thresh)
if sel.any():
# retrieve and return points
points = np.argwhere(sel)
#points = np.roll(points, 1, axis=1)
return points
else:
# otherwise return None to avoid having an empty list
return None
[docs]def plot_points(img, points, vminmax=(0,1), dims=None, invert=False, show=False):
'''Plot the detected points on the input image for checking.
Parameters:
img (np.ndarray): Image.
points (np.ndarray): Array containing the points.
vminmax (tuple): Tuple of two values for relative lower and upper cut off to display image.
dims (tuple): Tuple of dims to plot in dimensions.
invert (bool): Set to invert the image.
show (bool): Set to directly show the plot interactively.
Returns:
(np.ndarray): Image of the plot.
'''
try:
assert(isinstance(img, np.ndarray))
assert(isinstance(points, np.ndarray))
assert(points.shape[1] == 2)
assert(len(points.shape) == 2)
except:
raise TypeError('Something wrong with the input!')
fig = plt.figure()
ax = fig.add_subplot(111)
if invert:
cmap="Greys"
else:
cmap="gray"
if dims:
ax.imshow(img, cmap=cmap, vmin=np.min(img)+vminmax[0]*(np.max(img)-np.min(img)), vmax=np.min(img)+vminmax[1]*(np.max(img)-np.min(img)), extent=(np.min(dims[0][0]), np.max(dims[0][0]), np.max(dims[1][0]), np.min(dims[1][0])) )
ax.set_xlabel('{} {}'.format(dims[0][1], dims[0][2]))
ax.set_ylabel('{} {}'.format(dims[1][1], dims[1][2]))
ax.set_xlim((np.min(dims[0][0]), np.max(dims[0][0])))
ax.set_ylim((np.max(dims[1][0]), np.min(dims[1][0])))
else:
ax.imshow(img, cmap=cmap, vmin=np.min(img)+vminmax[0]*(np.max(img)-np.min(img)), vmax=np.min(img)+vminmax[1]*(np.max(img)-np.min(img)) )
ax.set_xlim((0,img.shape[1]-1))
ax.set_ylim((img.shape[0]-1,0))
ax.scatter(points[:,1], points[:,0], color='r', marker='o', facecolors='none')
if show:
try:
plt.show(block=False)
except:
pass
fig.canvas.draw()
plot = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')
plot = plot.reshape(fig.canvas.get_width_height()[::-1] + (3,))
return plot
[docs]def points_todim(points, dims):
'''Convert points from px coordinates to real dim.
Points are expected to be array indices for the first two dimensions in dims.
Parameters:
points (np.ndarray): Points to convert.
dims (tuple): Tuple of dimensions.
Returns:
(np.ndarray): Converted points.
'''
try:
# try to convert input to np.ndarray with 2 columns (necessary if only one entry provided)
points = np.reshape(np.array(points), (-1,2))
# check if enough dims availabel
assert(len(dims)>=2)
assert(len(dims[0])==3)
except:
raise TypeError('Something wrong with the input!')
# do the conversion by looking up thing in dimension vectors
points_d = np.array( [ dims[0][0][points[:,0]], dims[1][0][points[:,1]] ] ).transpose()
return points_d