Application of self-organising digital neural networks to attentive vision systems

Most image recognition systems treat each pixel in an image with dual significance (certainly in the early processing stages). Such an approach ignores one of the basic strengths of the human visual system. We attend to a very small region of the entire visual field at any one time. The ability to select important features-that is their identification and localisation-greatly reduces the processing needs of any vision system (animate or mechanical)