I'm reading a lot about vocabulary trees when it comes to fast queries for similar images or texts in big databases. But I couldn't find any good (easy to unserstand) description on what such an vocabulary tree is and how to build one out of features.