Facial recognition research originated from the fundamental task of distinguishing individuals based on their facial characteristics. Early researchers did exactly that by measuring distances between eyes, noses and mouths, but these geometric tricks fell apart as soon as a smile, tilt or shadow got in the way. So they began looking at the whole face, compressing images into “eigenfaces” that captured the most common patterns. These principal components were fast and easy but not very discerning, prompting the search for discriminative features with Fisherfaces and even higher‑order patterns with Independent Component Analysis. Others zoomed in on tiny patches, counting local binary patterns or filtering textures with Gabor wavelets to withstand changes in lighting and expression. This chapter traces that journey from hand‑measured landmarks to clever mathematical subspaces and texture codes. It also touches on iconic datasets like Labeled Faces in the Wild, explains what it means for a system to falsely accept or reject someone, and reflects on why these “classical” techniques still matter. Even though deep neural networks now steal the spotlight, the simplicity and transparency of these older methods keep them useful for teaching, benchmarking and environments with little data or limited hardware.