PerceptualHash is quite different on version 14.0.0 vs prior versions. Expected?

[ozziepeeps]

Hi, thanks for a great library!

I have observed that the PerceptualHash generated by version 14.0.0 is quite different to prior versions. I have created a repo that demonstrates what I'm seeing: https://github.com/ozziepeeps/MagickNET-perceptualhash-change

The repo includes a single (public domain) image asset. I have then written some unit tests and generated the perceptual hash on:

• Magick.NET-Q16-AnyCPU 7.16.0
• Magick.NET-Q16-AnyCPU 13.10.0
• Magick.NET-Q16-AnyCPU 14.0.0

All 3 package versions are present in the csproj for your convenience.

I observe that the PerceptualHash changes ever so slightly between 7.16.0 and 13.10.0, but the unit test HashDifference_MagickNET_Q16_AnyCPU_7_16_0_VS_MagickNET_Q16_AnyCPU_13_10_0 demonstrates that the difference is negligible < 0.00001.

However the unit test HashDifference_MagickNET_Q16_AnyCPU_13_10_0_VS_MagickNET_Q16_AnyCPU_14_0_0 demonstrates that the difference between 13.10.0 and 14.0.0 is much larger: 24.356.

I am not sure if this is intended. I've observed pretty stable / negligible changes across versions between 7.16.0 and 13.10.0 over my years of using the NuGet, but this one stood out so I wanted to highlight it just in case it wasn't intended.

Thanks!

[dlemstra]

Thanks for reaching out but this was an intended change and mentioned in the changelog of release 14.0.0: https://github.com/dlemstra/Magick.NET/releases/tag/14.0.0

• Changed the default colorspaces for PerceptualHash from sRGB and HCLp to XyY and HSB

You will get the "old" result if you do image.PerceptualHash(ColorSpace.sRGB, ColorSpace.HCLp) instead. This was an intended change because ImageMagick changed it's default implementation a while back. And with a major release I could finally make that change in this library.

[ozziepeeps]

Ah! I missed that when going through the release notes. My bad 👍 I will try your workaround for the "old" result in a few hours and report back here.

Whilst I have your attention, what is the theoretical maximum "difference" one can get from SumSquaredDistance? I don't recall ever being aware of or seeing a maximum value...

[dlemstra]

I don't know what the maximum value can be. I don't have the math expertise to give you an answer for that question.

[ozziepeeps]

No problem! I can confirm that the workaround you suggested works for me. I shall mark your reply as the answer. Thanks again!