PerceptualHash(sRGB, HCLp) output reverted to pre-14.7.0 values in 14.14.0... intentional?

[ozziepeeps]

Hi @dlemstra — following up on #1883.

After updating from 14.13.1 → 14.14.0 (ImageMagick 7.1.2-23 → 7.1.2-25), we've observed that image.PerceptualHash(ColorSpace.sRGB, ColorSpace.HCLp).ToString() produces output that is byte-for-byte identical to pre-14.7.0 values across all of our test images — meaning the algorithm change introduced in 14.7.0 appears to have been fully reverted.

In #1883, you described the 14.7.0 change as intentional bug fixes to ImageMagick's mathematical methods that "took several weeks to complete." The 14.14.0 release notes don't mention any perceptual hash changes.

Before we recompute our stored image hash library (which is substantial and was recomputed for the 14.7.0 change), we wanted to confirm:

1. Is this reversion in 14.14.0 intentional?
2. Should we expect the output to remain stable going forward on 14.14.0+?

For reference — one of our test images (49606_110.jpg) produced:

• 7.16.0: a8beb...621c48b420...8a7a261d19
• 14.7.0–14.13.x: a8beb...62ee062ee0...61d19 (the clamped-pattern values)
• 14.14.0: a8beb...621c48b420...8a7a261d19 (identical to 7.16.0)

[dlemstra]

This was an intentional change. I also noticed the same in my unit tests and asked if this was correct and they confirmed it was. And I was as surprised as you are. So yes you will need to update your hashes again but I cannot make any promises that it will not change in the future because they might change it again if they find a bug.

[ozziepeeps]

No problem sir. I'm glad we have the discussion as documentation at least :). I'll update my library of hashes, then.