How to reconstruct scenes from images using geometry and algebra, with applications to computer vision.
Hartly, in the US private sector, and Zisserman (U. of Oxford) describe techniques for reconstructing a real-world scene from several images of it, a basic problem in computer vision. The methods have been developed from projective geometry and photogrametry, and have the special feature of being uncalibrated, that is, it is not necessary to know or to compute the camera's internal parameters before getting an answer to the problem. They discuss the geometric principles, their algebraic representation in terms of camera projection matrices, the fundamental matrix, and the trifocal tensor. Annotation c. Book News, Inc., Portland, OR (booknews.com)