Images are central to computer vision, as they are the representation that we obtain from imaging devices (such as cameras; see Section 2.1). They provide us with a representation of the visual appearance of a scene (see Sections 2.2 and 2.3), which we can process to enhance certain features of interest, before we attempt to abstract information. Images generally exhibit some degree of noise (see Section 2.4) which can be attenuated by various simple image processing techniques (see Section 2.5).
A camera consists of a photosensitive image plane (which senses the amount of light that falls upon it), a housing which prevents stray light from falling onto the image plane and a lens in the housing which allows some light to fall onto the image plane in a controlled fashion (i.e. the light rays are focused onto the image plane by the lenses).
2.1.1 The Simple Pinhole Camera Model
One of the simplest, but reasonably realistic, models of a camera is the pinhole camera model in which the lens is instead treated as a simple pinhole (see Figure 2.1). All rays of light that hit the image plane must come through the pinhole of the camera in front of the photosensitive image plane. This is a simplification of most real imaging systems, as typically various parts of imaging systems (such as the lenses) introduce distortions into the resultant images. This basic model is extended to cope with these distortions in Section 5.6.