if I have an array like
z = np.random.random((41,61,106))
y,x=np.mgrid[slice(0,61, 1),slice(0,106, 1)]
z=z[_num,x,y]]
Who I can capture the 10 maximum and 10 minimum and mask the rest of the values in a array z.
And, it is posible take the midle of the max and the min values an put another 10 values.
middle=sorted[(len(sorted)/2)-5:(len(sorted)/2)+5]
print middle
mask = ma.masked_inside(z,sorted[10],sorted[-10],middle)
print mask
I do not quite understand the code snippet. To answer your first question:
import numpy as np
import numpy.ma as ma
z = np.random.random((10,10))
sorted = np.sort(z,axis=None)
mask = ma.masked_inside(z,sorted[10],sorted[-10])
concerning your second question you might also consider concatenating the conditions
ma.masked_where( ((z<sorted[sorted.size/2-5]) |
(z>sorted[sorted.size/2+4])) &
((z>sorted[10]) & (z<sorted[-10])),z)
If you only need a few of the items in an array to be sorted, with numpy >= 1.8 it is more efficient to use np.partition than np.sort:
In [6]: z = np.random.rand(61, 106)
In [7]: %timeit np.sort(z, axis=None)
1000 loops, best of 3: 413 µs per loop
In [8]: %%timeit
...: n = z.size
...: y = np.partition(z, (10, n//2 - 5, n//2 + 5, -10), axis=None)
...: y[:10].sort()
...: y[n//2 - 5:n//2 + 5].sort()
...: y[-10:].sort()
...:
10000 loops, best of 3: 143 µs per loop
