Masked pixels appear in segmentation map

[ysBach]

The docs says

Masking a pixel is equivalent to setting data to zero and noise (if present) to infinity.

A test case to demonstrate this:

import numpy as np
import sep

mask = np.zeros((7, 7)).astype(bool)
mask[3, 3:] = True
data = np.ones((7, 7))
data[2:5, 2:5] += 20
obj, seg = sep.extract(data, thresh=1,
                       mask=mask, 
                       segmentation_map=True)
print(data)
print(mask.astype(int))
print(seg)
print(obj["x"], obj["y"])

# image
[[ 1.  1.  1.  1.  1.  1.  1.]
 [ 1.  1.  1.  1.  1.  1.  1.]
 [ 1.  1. 21. 21. 21.  1.  1.]
 [ 1.  1. 21. 21. 21.  1.  1.]
 [ 1.  1. 21. 21. 21.  1.  1.]
 [ 1.  1.  1.  1.  1.  1.  1.]
 [ 1.  1.  1.  1.  1.  1.  1.]]
# mask
[[0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 1 1 1 1]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]]
# Segmap
[[0 0 0 0 0 0 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 0 0 0 0 0 0]]
# Final x, y
[2.8595679] [3.]

The obj is correctly calculated after masking the pixels. (∵ obj["x"] != 3.0)
However, seg includes masked pixels, too.
I think this might be intentional, because the pixels used for extract can be retrived by seg & ~mask, while the additional segmentation information is given to the user. (I am not sure what is SExtractor's default, as I have never used it)
However, if that's the case, I think the documentation needs a bit more information.

[PJ-Watson]

The docs says

Masking a pixel is equivalent to setting data to zero and noise (if present) to infinity.

A test case to demonstrate this:

import numpy as np
import sep

mask = np.zeros((7, 7)).astype(bool)
mask[3, 3:] = True
data = np.ones((7, 7))
data[2:5, 2:5] += 20
obj, seg = sep.extract(data, thresh=1,
                       mask=mask, 
                       segmentation_map=True)
print(data)
print(mask.astype(int))
print(seg)
print(obj["x"], obj["y"])

The obj is correctly calculated after masking the pixels. (∵ obj["x"] != 3.0) However, seg includes masked pixels, too. I think this might be intentional, because the pixels used for extract can be retrived by seg & ~mask, while the additional segmentation information is given to the user. (I am not sure what is SExtractor's default, as I have never used it) However, if that's the case, I think the documentation needs a bit more information.

This is actually quite an interesting issue - thanks for pointing this out! This is the intended behaviour (there is no equivalent in SExtractor to my knowledge), and stems from the order in which operations are applied. The mask is applied before the default filtering, which becomes clearer when turning this off (filter_kernel=None):

Default filter

# image
[[ 1.  1.  1.  1.  1.  1.  1.]
 [ 1.  1.  1.  1.  1.  1.  1.]
 [ 1.  1. 21. 21. 21.  1.  1.]
 [ 1.  1. 21. 21. 21.  1.  1.]
 [ 1.  1. 21. 21. 21.  1.  1.]
 [ 1.  1.  1.  1.  1.  1.  1.]
 [ 1.  1.  1.  1.  1.  1.  1.]]
# mask
[[0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 1 1 1 1]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]]
# Segmap
[[0 0 0 0 0 0 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 1 1 1 1 1 0]
 [0 0 0 0 0 0 0]]
# Final x, y
[2.8595679] [3.]

`filter_kernel=None`

# image
[[ 1.  1.  1.  1.  1.  1.  1.]
[ 1.  1.  1.  1.  1.  1.  1.]
[ 1.  1. 21. 21. 21.  1.  1.]
[ 1.  1. 21. 21. 21.  1.  1.]
[ 1.  1. 21. 21. 21.  1.  1.]
[ 1.  1.  1.  1.  1.  1.  1.]
[ 1.  1.  1.  1.  1.  1.  1.]]
# mask
[[0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 1 1 1 1]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]]
# Segmap
[[0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]
 [0 0 1 1 1 0 0]
 [0 0 1 0 0 0 0]
 [0 0 1 1 1 0 0]
 [0 0 0 0 0 0 0]
 [0 0 0 0 0 0 0]]
# Final x, y
[2.85714286] [3.]

Or to show the difference graphically, this is how the input data changes when a mask is given:

Figure

The quantities are measured from these masked (and convolved) images, so it's not correct that the extracted pixels can be retrieved from seg & ~mask. This can be checked by comparing the number of pixels detected above the threshold - either tnpix, or npix - with and without the mask.