How is this data filter technique called

A while ago I read about a data filter technique or algorithm that could quickly determine if an object was not similar to an object in a group of objects. It was explained as having multiple overlaying stencils where each object is one stencil. The object (imagined as a cut out piece of paper) is placed underneath these group of stencils.

If some of the object is visible, than you know for certain that it is not similar to one of the other objects/stencils; if the object was not visible, there is a change it is similar to one of the objects, but that is not certain.

This is used for fast determining negatives, while a possible positive takes some more time (it must be compared to all of the other stencils to proof similarity).

I can not remember the name of this technique and I want to read more about it. How is it called?