Please could you add tests for this?

PIL/ImageOps.py (1.7%): Missing lines 150-156,158-160,162-164,167-169,171-175,177-180,182-186,188-191,193-200,202-204,206,208-218
-------------
Total:   58 lines
Missing: 57 lines
Coverage: 1%

https://travis-ci.org/python-pillow/Pillow/jobs/64760914#L3778

Docs would be very welcome too.

Not quite sure what you're looking for in terms of docs, but I added "while preserving image tone" to a copy of the autocontrast docstring.

Changes Unknown when pulling 93e6c08 on broxeph:master into * on python-pillow:master*.

The tests that were added do show that it doesn't crash, but don't validate correctness.

Possible Additional tests:

• GS ramp of 256 greys, cutoff n% do we get the expected number of full white/black pixels.
• Opposite ramp in 2 color channels, constant in the other.
• Single color image, grey, black, white. (or single channel.)
• Synthetic image with ramp in one channel +constant value + light noise in the other channels.

I've added some tests here (wiredfool@bb3ad27) for correctness of implementation, and I'm seeing some really strange behavior for running against a constant color image.

If all pixels are white, it returns a white image. (127, 255,0) returns the same. But (127,127,127) returns an image that's (127, 255, 255), and an all black image returns (0,255,255). I don't think that this is intended behavior.

I'll close this because there's been no progress in 3-4 years. Ping if you'd like to re-open it, or create a new PR. Thanks!

An alternative to this, #5350, has been merged