Multiple view geometry in computer vision. (English) Zbl 0956.68149

Hartley, Richard; Zisserman, Andrew

Cambridge: Cambridge University Press. xvi, 607 p. (2000).

This book is concerned with the relations between two or more linear images of spatial objects, and the reconstruction of these spatial objects. The main idea is the use of ‘epipolar geometry’: For any two linear (in the sense of projective geometry) images of space there is a certain singular bilinear \(F(x_1,x_2)\) which takes as arguments homogeneous coordinate vectors \(x_i\) image points and which is zero if \(x_1\), \(x_2\) are the images of the same point. Generalizations to more than two images are straightforward. The book is apparently written for the ‘computer vision’ community. It contains a very readable and detailed discussion of various problems. It uses the language of analytic projective geometry and the reader should have few problems in applying the formulae and algorithms directly. As ‘computer vision’ is a new subject, the older literatur is consequently ignored, with a few exceptions, but this does not diminish the value of this book.

Reviewer: Johannes Wallner (Wien)

MSC:

68U10	Computing methodologies for image processing
51N15	Projective analytic geometry
68-01	Introductory exposition (textbooks, tutorial papers, etc.) pertaining to computer science

Keywords:

computer vision; epipolar geometry; image reconstruction