A structural model of the human face (graphics, animation, object representation)

This dissertation examines the problems of representing and animating complex, highly connected objects. Systems such as these are difficult to describe and animate due to the richness of actions and physical interconnections. A strategy of partitioning the objects, actions, and application algorithm is proposed which allows complete independence of definition of these three classes. A simple animator has been implemented for a gear-and-axle universe which is capable of detecting inconsistent cyclic systems while correctly animating consistent ones. An object of proven interest, the human face, is then examined. The face itself consists of a moderately large number of definable regions of expression and a predefined set of performable actions. An action in one of these regions may or may not cause other changes, both in the state of the region and in other adjacent regions. OASIS/F, using the object/action partition, was implemented. This representation of the face is capable of animating the face in a realistic manner, preserving the structural properties and idiosyncracies of the natural human face.