Image Feature Location in Multi-Resolution Images Using a Hierarchy of Multilayer Perceptrons

This chapter describes work undertaken at BT Laboratories to produce a system, based on the use of neural network feature detectors, to robustly locate and track features in digital image sequences. We have concentrated on the location of eyes and mouths in human head-and-shoulders images, as described by Nightingale in the introduction to this part of the book. However the techniques described should be applicable to determining the position of localised features in general images.