Just a behaviour note... about jimp HOT 4 CLOSED

Yes. I think that raises a question about how sensitive you expect it to be.

I also think it's a question of the right tool for the job:

When I was testing it, I used a cut-off of 15%, so 2% would be considered a match. And they are, if you are testing for similarity.

On the other hand, if you are testing for difference then the image.diff() method is probably the better tool.

from jimp.

I meant an additional 2%, I was verifiyng how many additional changes to an image would cause the threshold (I currently use 12%) to be overcome...

On the other hand, I don't completely understand the distinction you make between similarity and difference... Ultimately, they are just two different algorithms (one from 'pHash' and one from 'PixelMatch') to measure the same entity... Or am I wrong?

from jimp.

They are two different algorithms though.

If pHash and PixelMatch were people, pHash would scrunch up its eyes and say that a cat looks like a dog. PixelMatch, on the other hand, would carefully detail all the things about a cat that a different to a dog within a given threshold.

Both are right. pHash is good at finding similarity. PixelMatch is good at finding difference.

from jimp.

I see... Your metaphor was great! :-) Now I clearly understand what you meant...
Now I'm even more convinced I'll use pHash (Jimp.hash()) to find similarities on my images...
Thanks a lot, Oliver!

from jimp.