In this chapter, we learnt how to convert our image into a feature vector, or a sequence of values. Specifically, we looked at one particular feature descriptor: the local binary pattern (or, LBP) operator. Apart from the traditional LBP formulation, we also saw some common variants that are used. An important takeaway from this chapter is that the LBP captures the "texture" of the input image. We also looked at how such texture information can help us capture the subtle variations present in facial images.
Local binary pattern is just one of the many possible feature descriptors that have been proposed in Computer Vision literature. Some other examples include Histogram of Oriented Gradients (HoG), SIFT and SURF. HoG uses the concept of image derivatives whereas SIFT and SURF are much more sophisticated feature descriptors (they are patented algorithms as well). In fact, similar in spirit to LBP, we have yet another feature descriptor called local ternary patterns (LTP). Instead...