Skip to content

Allow array alpha for imshow #8183

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
tillahoffmann wants to merge 10 commits into from
Closed

Allow array alpha for imshow #8183

Changes from 1 commit
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
157fc86
Allow array alpha for imshow.
tillahoffmann Mar 2, 2017
8916fa9
Remove unnecessary logic branch.
tillahoffmann Mar 2, 2017
fd0ed02
Fix PEP8.
tillahoffmann Mar 3, 2017
c9c96ab
Only restore alpha channel if `_array_alpha` is given.
tillahoffmann Mar 3, 2017
f04266b
Merge remote-tracking branch 'upstream/master' into arrayalpha
tillahoffmann Mar 3, 2017
f6e788a
Rerender test images.
tillahoffmann Mar 3, 2017
efc8b74
Add test for array alpha.
tillahoffmann Mar 7, 2017
b9bce13
Update docstring.
tillahoffmann Mar 7, 2017
c5437e5
Merge remote-tracking branch 'upstream/master' into arrayalpha
tillahoffmann Mar 7, 2017
fdbc951
Trigger CI
tillahoffmann Mar 7, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 18 additions & 3 deletions lib/matplotlib/image.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -237,6 +237,7 @@ def __init__(self, ax,
self.axes = ax

self._imcache = None
self._array_alpha = None

self.update(kwargs)

Expand All @@ -260,7 +261,11 @@ def set_alpha(self, alpha):

ACCEPTS: float
"""
martist.Artist.set_alpha(self, alpha)
if np.isscalar(alpha):
martist.Artist.set_alpha(self, alpha)
else:
self._array_alpha = alpha
martist.Artist.set_alpha(self, 1.0)
self._imcache = None

def changed(self):
Expand Down Expand Up @@ -409,6 +414,10 @@ def _make_image(self, A, in_bbox, out_bbox, clip_bbox, magnification=1.0,
if alpha is None:
alpha = 1.0

# Apply the array alpha if provided
if self._array_alpha is not None:
A[..., 3] *= self._array_alpha

_image.resample(
A, output, t, _interpd_[self.get_interpolation()],
self.get_resample(), alpha,
Expand All @@ -419,14 +428,18 @@ def _make_image(self, A, in_bbox, out_bbox, clip_bbox, magnification=1.0,
# colormapping works correctly
hid_output = output
output = np.ma.masked_array(
hid_output[..., 0], hid_output[..., 3] < 0.5)
hid_output[..., 0], hid_output[..., 3] == 0)

@tacaswell tacaswell Mar 2, 2017

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is going to break a test, but may fix an issue that @QuLogic found with imshow after #8024

@tillahoffmann tillahoffmann Mar 3, 2017

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, this line is indeed causing the problem. In particular, the alpha channel is interpolated near the boundary of an image. The < 0.5 implementation will set pixels that are "more than half transparent" to fully transparent whereas the == 0 implementation only treats them as masked if they are fully transparent.

In short, the boundaries in a few tests differ slightly.

rotate_image-failed-diff

I believe the == 0 implementation is preferable because it effectively provides anti aliasing on the boundary by restoring the alpha channel. Here is a zoomed-in example.

screen shot 2017-03-03 at 15 03 01
screen shot 2017-03-03 at 15 03 06

@tillahoffmann tillahoffmann Mar 3, 2017

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hm, but this causes trouble if there are bad data in image. For example, bad values in one of the tests are shown as blue. If we restore the alpha channel, all bad blue values will become transparent. As a compromise, I have rerendered the problematic images (changes only occur at the boundary) and I only restore the alpha channel if _array_alpha is given.

# relabel under data
output[hid_output[..., 1] > .5] = -1
# relabel over data
output[hid_output[..., 2] > .5] = 2

output = self.to_rgba(output, bytes=True, norm=False)

# Restore the alpha channel
if created_rgba_mask:
output[..., 3] = (output[..., 3] * hid_output[..., 3]).astype(output.dtype)

# Apply alpha *after* if the input was greyscale without a mask
if A.ndim == 2 or created_rgba_mask:
alpha = self.get_alpha()
Expand Down Expand Up @@ -495,7 +508,9 @@ def draw(self, renderer, *args, **kwargs):
# actually render the image.
gc = renderer.new_gc()
self._set_gc_clip(gc)
gc.set_alpha(self.get_alpha())
# If the alpha channel is not a sclar, just use full opacity for the artist
alpha = self.get_alpha()
gc.set_alpha(alpha if np.isscalar(alpha) else 1.0)
gc.set_url(self.get_url())
gc.set_gid(self.get_gid())

Expand Down