This book is about machine vision research. The chapters are slightly abridged Ph. D. theses or other similarly major works, all from the M. I. T. Artificial Intelligence Laboratory. The central themes are as follows: B. Horn demonstrates how the coordinates of a curved surface can be derived from surface intensity variation. Y. Shirai shows how knowledge about blocks can help drive a line finder over an intensity array in order to find objects and create a line drawing. P. Waltz confirms that the structure of a line drawing forces conclusions about which lines separate bodies and about which lines are shadows or cracks. P. H. Winston explains how a machine can quickly learn the essence of concepts involving simple configurations like that of an ″arch.″ M. Minsky discusses a theory of abstract symbolic representation designed to cope with movement in and reasoning about three-dimensional space. The theories of Horn, Shirai, Waltz, and Winston are all supported by experiments with program implementations. Minsky's theory is ready to be filled out and refined through such implementations. References are provided at the end of each chapter, in addition to a general bibliography.
