Autonomous motion vision

摘要

Earlier, the author (1991) introduced a direct method calledfixation for the recovery of shape and motion in the general case. Themethod uses neither feature correspondence nor optical flow. Instead, itdirectly employs the spatio-temporal gradients of images brightnesses.The present paper reports the experimental results of applying some ofthe fixation algorithms to a sequence of real images where the motion isa combination of translation and rotation. Techniques for autonomouschoice of parameters which result in good estimates for important motionparameters are also described