Distance comparing algorithm might be wrong

[BigPanda97]

From my understanding of your code, it looks like you are using the mean square error to compare two hashes. But in this document, one of the authors of PhotoDNA states that: “Finally, we compute the similarity of two hashes as the Euclidean distance between two feature vectors, with distances below a specified threshold qualifying as a match.”

So, it looks like the official implementations use the Euclidean distance instead of the mean square error.