§ 1. Introduction
This chapter introduces the basic geometric concepts of multiple-view geometry in computer vision. The focus is on geometric models of perspective cameras, and the constraints and properties such models generate when multiple cameras observe the same 3-D scene.
Geometric vision is an important and well-studied part of computer vision. A wealth of useful results has been achieved in the last 15 years and has been reported in comprehensive monographies, e.g., (Faugeras,1993; Hartley and Zisserman,2003; Faugeras and Luong,2001), a sign of maturity for a research subject.
It is worth reminding the reader that geometry is an important but not the only important aspect of computer vision, and in particular of multi-view vision. The information brought by each image pixel is twofold: its position and its colour (or brightness, for a monochrome image). Ultimately, each computer vision system must start with brightness values, and, to smaller or greater depth, link such values to the 3-D world.