Fuzzy pixel color comparison

[kaczymuczy]

Description

Introduce an option to match colors that are close to the target color in a grayscale space when comparing pixels with the LayoutComparator.
The comparison can now run in two modes:

• the standard mode (nothing changed),
• "fuzz mode" - the baseline pattern and new screenshot images are converted to grayscale and the comparison takes into account a "fuzz" - a value that describes how different (how big is the distance) the pattern and screenshot pixels can be in order to consider them equal.
  In other words - a naive implementation of https://imagemagick.org/script/command-line-options.php#fuzz

Motivation and Context

This option allows to filter out image rendering issues like AEM image rendition differences not visible to the human eye.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)

Checklist:

• My code follows the code style of this project.
• I have reviewed (and updated if needed) the documentation regarding this change

---

I hereby agree to the terms of the AET Contributor License Agreement.

[kaczymuczy]

Testing on a working AET instance not done yet.

[kaczymuczy]

Documentation not done yet.

[kaczymuczy]

Testing on a working AET instance - done, works fine.

[kaczymuczy]

Documentation done

[wiiitek]

We could add some slightly different images (lie those generated by AEM renditions engines) to our integration tests with example of how to use fuzz parameter in test suite.
What do you think about such idea?

[kaczymuczy]

#507 (comment)
The unit tests pretty much cover everything there is to test for this feature and the documentation covers the usage example. I don't think extra integration tests would bring more value - you believe this is necessary?

[wiiitek]

#507 (comment)
Thank you for your response!

I don't think they are really necessary. And also I don't know what is the role of integration tests currently.

I have looked into sample files for integration tests - they are taken from the real-life use cases, right?

[plutasnyy]

I like this feature very much! Problems with images were really really annoying. I think a bit about users - if a test is marked as a fail user doesn't know if it was because of threshold or fuzzy. I also think that using it together is not the best way so I wouldn't change anything.
The second thing that I wouldn't know is how big is the difference between some values of fuzzy, maybe some comparison of noised images that pass a test with the fuzzy value equal to 5% 10% 15% etc could be helpful?
Please update changelog :D
Good job!

[kaczymuczy]

#507 (comment)
Yes, the samples are built from a number of real life cases.

Changelog updated

So, what exactly am I required to do to have this merged? I would love to use this as soon as possible. :D

[plutasnyy]

From my side you have approval. Good luck ;)