Basic Contents

Computer Vision ... (Published 1992)

Simulated Human Vision..... Ian Overington

Location: Eastbourne. UK
ianoverington@simulatedvision.co.uk ............ www.simulatedvision.co.uk

9.9.3. Global motion.
9.9.4. Images having global mismatch together with local motion.
9.9.5. 3-D flow disruption boundaries.
9.9.6. Focus of expansion.
9.10. Processing of extended sequences.
9.11. Complete flow analysis and the X-point algorithm.
9.12. Conclusions.

Chapter 10. Stereo.


There are numerous situations, both in industrial inspection and in other engineering tasks, where
the concepts of stereopsis are seen by many to be potentially a valuable source of depth (range)
information. This is largely because of the widespread 'lay' knowledge of the apparent values of
human stereo vision, and because of the proven usefulness of optical (visual) rangefinders utilising
stereo concepts. There has been a great deal of activity in recent years towards the automatic
sensing of stereo depth data in stereo pairs of images, particularly by Marr and colleagues and by
Mayhew, Frisby and colleagues. Such studies have been plagued, in particular, by the so-called
'correspondence' problem - that is the need to relate particular features in one image of a stereo
pair to the correct features in the other image (see also Chapter 9.2 for discussion). The result of
this has been the development of elaborate algorithms, which explore characteristics of a region on
one image around the location of a selected feature on the other. Whilst these algorithms are
impressive in their function, they are nevertheless rather cumbersome. Also practical
implementation tends to be limited to the determination of a strictly finite number of stereo features in
any one stereo pair, the stereo resolution being relatively low. It is suggested that such cumbersome