We have spent a good amount of time working with faces. Over the course of the last couple of chapters, we have learnt how to detect faces in a given image and how to align a set of faces so that they are of the same size and none of the faces have a head-tilt either to the right or to the left. We also mentioned that such a pre-processing step is necessary to enhance the performance of machine learning algorithms. We did a very brief discourse on image classification systems when we were discussing face detectors in Chapter 6, Face Detection Using OpenCV. We introduced the notion of features (specifically, Haar features). This chapter, along with the next, is going to be an in-depth analysis of these concepts.
When we introduced the concept of digital images right at the beginning of the very first chapter, we said that an image is represented by its set of pixel values. During the formative chapters, we played around with these pixel values to perform...