ABSTRACT
Vision is one of the five human senses, and it is the one that carries the richest information content. Humans rely on vision for almost all aspects of living, including particularly locomotion, searching for sustenance, and manufacturing myriad edifices, devices, and food products. Thus, it is natural to try to endow man-made machines with vision, with a view to adding to their utility and capabilities. Indeed, it is commonplace to envisage robots as machines that are intelligent and able to move around and act like humans. The necessary transition from biological vision to machine and robot vision hides the fact that vision is a complex ability that has evolved over millions of years to be highly powerful, and apparently effortless and instantaneous, in operation. However, it was discovered long ago that vision is actually quite difficult to engineer and proceeds in two stages that are often confused-seeing and perceiving. The eyes see and pass on basic picture information to the human brain, while the brain itself carries out the perception or detailed understanding of the scene being viewed; this perception is in turn mediated by complex processes of comparison with the huge database of information about the real world that is stored in the brain. By definition, biological and machine vision systems can only carry out processes that are possible, and the scientific study of what is possible is called
computer vision
. It is, then, the application of computer vision to real-world tasks and problems that is the domain of machine vision.
