Same result for different images

I really like this library. But I need some improvement. Like in this example. The hash is the same, but the images are different. I need more accuracy and certainty that the pictures are the same.

How to do it?

Start by reproducing the reported case using the image linked in the issue and inspect how the library generates and compares perceptual hashes. Determine why different images produce the same result and define what level of accuracy or certainty is expected before proposing a change. Done means the reported images no longer compare as identical without breaking valid matching behavior.

Written by the indexing model from the issue text.