This paper presents this emerging computational view of visual perception, discusses some early work in the field in its context, and puts forward current thoughts on the overall organization and operation of a general-purpose computer vision system, synthesizing recent theoretical and experimental results.
Research is beginning to uncover fundamental computational principles underlying vision that apply equally to artificial and natural systems. These principles provide insights into the limitations of early machine vision systems and lay a foundation for building future systems capable of high performance in a broad range of visual domains. We present this emerging computational view of visual perception, discuss some early work in the field in its context, and put forward current thoughts on the overall organization and operation of a general-purpose computer vision system, synthesizing recent theoretical and experimental results.