Hearing recently from a friend that in the new OS Windows there will be image authorization decided to try to implement the same system but under the Web interface. After 2 hours of experiments, the quick registration turned out to be quite satisfactory, requiring only the entry of a username (login).
You can see here . During authorization, previously entered coordinates x + -10px y + -10px are used. Also required to comply with the sequence of pressing points.
It seems to me that there is no sense in using custom projects. users will either forget the specified points, or most users will have the same points, which makes such authorization not secure. ')
Sources
UPD: Fixed a bug with Firefox browser. He has a separate system for obtaining coordinates from the image.