Investigating Face Recognition with an Image Processing Computer

Abstract

All of us participating in this conference will, at some time or other, have ruminated deeply about faces and how we recognise them. The human visual recognition process is so astonishingly effortless, however, that most people take it entirely for granted, without questioning the extended complexity of neural processing that must be involved. So much so that it is sometimes difficult for people to realise just how complex is the process of visual perception, quite apart from the additional burden of recognition. Indeed, the Nobel Prizewinner Francis Crick (1979) expressed the problem well when he said “Few people realise what an astonishing achievement it is to be able to see at all”. One of the aims of my group at RARDE is to attempt an understanding of visual recognition, in order that it may be modelled and, ultimately, simulated by computers and electro-optic imaging systems. Clearly, an early stage in the development of modelling must involve the collection of relevant data in several different fields of human endeavour, such as physical optics, psychology and neurophysiology. The decision to use faces as recognition targets was based on their essential ubiquity and familiarity; observers would not require special training to recognise them.