Learning OpenCV

Chapter 11. Camera Models and Calibration

Vision begins with the detection of light from the world. That light begins as rays emanating from some source (e.g., a light bulb or the sun), which then travels through space until striking some object. When that light strikes the object, much of the light is absorbed, and what is not absorbed we perceive as the color of the light. Reflected light that makes its way to our eye (or our camera) is collected on our retina (or our imager). The geometry of this arrangement—particularly of the ray's travel from the object, through the lens in our eye or camera, and to the retina or imager—is of particular importance to practical computer vision.

A simple but useful model of how this happens is the pinhole camera model. ^[164] A pinhole is an imaginary wall with a tiny hole in the center that blocks all rays except those passing through the tiny aperture in the center. In this chapter, we will start with a pinhole camera model to get a handle on the basic geometry of projecting rays. Unfortunately, a real pinhole is not a very good way to make images because it does not gather enough light for rapid exposure. This is why our eyes and cameras use lenses to gather more light than what would be available at a single point. The downside, however, is that gathering more light with a lens not only forces us to move beyond the simple geometry of the pinhole model but also introduces distortions from the lens itself.

In this chapter we will learn ...