From surfaces to objects: Computer vision and three dimensional scene analysis

A unified approach to the theory and practice of computer vision. Presents a model-based, 3-dimensional scene analysis that combines surface patches segmented from the 3-dimensional scene description; surface-patch-based object models; a hierarchy of representations, models, and recognitions; a distributed-network-based model invocation process; and a knowledge-based model matcher. Describes the model-independent scene analysis, and how objects are represented and selected, and shows how to locate, verify, and understand a known object given its geometric model.