The problem of detecting a face in an image can be posed in a simpler way. We could iterate the whole image through several smaller windows and create a classifier that will tell whether a window is a face or not. The windows that correctly identify the face will be the coordinates of face detection.

Now, what exactly is a classifier and how can it be built? In machine learning, the problem of classification has been deeply explored and it is posed as the identification of which of the set of categories a given observation belongs to, based on a previously trained set of known category memberships. This could be something like if a given image belongs to the banana, apple, or grape category, for instance, in a fruit classification application. In the case of face detection, there are two categories—face and non-face.

This section describes a meta-algorithm, which is basically a templated algorithm to create a strong classifier using a set of weak learners. These weak learners...