Skip to content

Restore special casing of grayscale images in PostScript backend.#12871

Closed
anntzer wants to merge 1 commit intomatplotlib:masterfrom
anntzer:ps-grayscale
Closed

Restore special casing of grayscale images in PostScript backend.#12871
anntzer wants to merge 1 commit intomatplotlib:masterfrom
anntzer:ps-grayscale

Conversation

@anntzer
Copy link
Copy Markdown
Contributor

@anntzer anntzer commented Nov 23, 2018

This divides the filesize by ~3 by not storing three identical channels
(e.g. compare the size of

imshow(np.random.rand(255, 255), cmap="gray")
gca().set_axis_off()
savefig("test.ps")

before and after the patch).

Looks like some kind of similar functionality was present before mpl2.0, though with some weird magic factors (https://github.com/matplotlib/matplotlib/pull/5718/files#diff-b465967f537465fec4960b152fa49cffL415), and I stumbled on some vestigial leftovers of it (https://github.com/matplotlib/matplotlib/pull/5718/files#diff-b465967f537465fec4960b152fa49cffL457 where the "image" command comes from)...

No hard numbers but I don't think the performance cost (on non-grayscale images) really matters as we're going to iterate through the array just after to serialize it anyways.

Probably the same thing can be done for the pdf backend -- see grayscale parameter of _writeImg.

PR Summary

PR Checklist

  • Has Pytest style unit tests
  • Code is Flake 8 compliant
  • New features are documented, with examples if plot related
  • Documentation is sphinx and numpydoc compliant
  • Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
  • Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

@anntzer
Restore special casing of grayscale images in PostScript backend.
c650993 
This divides the filesize by ~3 by not storing three identical channels
(e.g. compare the size of
```
imshow(np.random.rand(255, 255), cmap="gray")
gca().set_axis_off()
savefig("test.ps")
```
before and after the patch).
@jklymak
Copy link
Copy Markdown
Member

jklymak commented Nov 23, 2018

Does this get tested somehow? Otherwise makes sense to me, unless the performance hit is significant, which I doubt as well...

@QuLogic
Copy link
Copy Markdown
Member

QuLogic commented Nov 24, 2018

Those factors are luminance conversions.

@anntzer
Copy link
Copy Markdown
Contributor Author

anntzer commented Nov 24, 2018

Gah, the obvious tests

@check_figures_equal(extensions=["eps"])
def test_grayscale_compression(fig_test, fig_ref):
    # Fully opaque, grayscale images are saved as single channel.  Compare
    # with a non-fully opaque image (by making the an all-white pixel
    # transparent...) which is saved as three different channels.
    t = np.arange(256, dtype="u1").reshape((16, 16))
    rgba = np.stack([t, t, t, np.ones_like(t, "u1") * 0xff], 2)
    fig_ref.subplots().imshow(rgba)
    rgba[-1, -1] = 0
    fig_test.subplots().imshow(rgba)

or

@check_figures_equal(extensions=["eps"])
def test_grayscale_compression(fig_test, fig_ref):
    # Fully opaque, grayscale images are saved as single channel.  Compare with
    # a non-fully opaque image which is saved as three different channels.
    test_data = np.arange(100).reshape((10, 10))
    fig_ref.subplots().imshow(test_data, cmap="gray")
    fig_test.subplots().imshow(test_data, cmap="gray", alpha=.5)

fail (visibly); I guess there's something about gamma correction being applied in one case and not the other by the renderer...

@anntzer anntzer self-assigned this Nov 26, 2018
@anntzer anntzer closed this Nov 26, 2018
@anntzer anntzer mentioned this pull request Jan 4, 2019
Merged
6 tasks
@anntzer anntzer mentioned this pull request Sep 3, 2019
Merged
6 tasks
@anntzer anntzer deleted the ps-grayscale branch March 20, 2022 13:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

@anntzer anntzer

Labels

backend: ps

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@anntzer @jklymak @QuLogic