A large amount of information about the world we live in is supplied by our visual systems. Humans perform the task of vision effortlessly without being aware of this complex process. Hierarchical structures permit successful examination of such intricate operations.
This book investigates hierarchical-structured neural networks for vision and image processing tasks and proposes various new neural network models for that purpose. It exploits the capabilities of hierarchical neural networks in a systematic way by considering the similarities to hierarchical structures already in use by computer vision researchers. All issues of hierarchical neural networks are treated in considerable detail; that is, the structure of the network, the representation issue, and learning mechanisms are analyzed theoretically as well as experimentally. Considering the similarity between conventional vision algorithms and hierarchical neural networks not only allows a transfer of knowledge between these two fields, but also gives voice to many new algorithms.
"The experiments reported in the book are highly informative, in several cases innovative, and always well-thought. The flow of the ideas is smooth and coherent. Overall, the book is an interesting and useful monograph that could be recommended to everybody pursuing studies at the junction of neurocomputing and hierarchical image processing."
Contents: Preface. Introduction. Networks Versus Pyramids. Structural Considerations. Contents and Representation. Processing. Learning Pyramids. Hierarchical Learning. Conclusion and Outlook.