This chapter explores a hierarchy of feature representations starting with classic, hand-crafted features. It presents a few classic feature representations that illustrate a wide variety of hand-crafted feature extractors. The chapter describes important features that characterize the color distribution in an image in a compact and efficient manner. It reviews some of the important latent feature types and the underlying extraction processes. Direct feature extraction has its own advantages. It is (usually) computationally efficient, easy to implement and interpret, and allows inclusion of prior knowledge. Images are the most common form of visual data in many visual computing applications. The chapter provides an example of a fundus image where the color spectrum is predominantly red and thus conventional binning strategies that work well in a natural setting fail there. Principal component analysis employs orthogonal transformations to convert a set of correlated variables into a set of linearly uncorrelated variables.