Except for the excellent typography, this volume is a typical book extracted (presumably) from a Ph.D. dissertation. It contains philosophistical (sic) ruminations and some humorous asides. It expands to 182 pages a fairly simple and noteworthy idea. That idea is simply that
it is possible to recover depth information from two images acquired at disparate locations in the environment using a sensor with fixed orientation with respect to the axis of motion.… [Then] the depth recovery problem is reduced to determining the disparity &dgr;.
This may sound like normal stereophotogrammetry (stereo) using a single camera, but it is not. Like stereo, the algorithm requires points of steep optical gradient in the scene, but the way the information is used is different. The correspondence problem is never solved. Instead, a relation is found between the temporal intensity gradient--the optical flow between successive frames--and the spatial intensity gradient. This relation gives the required disparity information. Not incidentally, a further difference from stereo is that axial motion as well as lateral motion will work.
The scheme was verified by having a robot navigate a room with various obstacles, including objects partially occluded at various points on the trajectory. Some of the layouts were taxing.