MOTION FROM FIXATION

We study the problem of estimating rigid motion from a sequence of mon ocular perspective images obtained by navigating around an object whil e fixating a particular feature-point. The motivation comes from the m echanics of the human eye, which either pursues smoothly some fixation point in the scene, or 'saccades' between different fixation points. In particular, we are interested in understanding whether fixation hel ps the process of estimating motion in the sense that it makes it more robust, better conditioned or simpler to solve. We cast the problem i n the framework of 'Epipolar geometry', and propose a filter based upo n an implicit dynamical model for recursively estimating motion under the fixation constraint. This allows us to compare directly the qualit y of the estimates of motion obtained by imposing the fixation constra int against the estimates obtained assuming a general rigid motion sim ply by changing the geometry of the parameter space, while maintaining the same structure of the recursive estimator. We also present a clos ed-form static solution from two views, a recursive estimator of the r elative attitude between the viewer and the scene and assess how the e stimates degrade in the presence of disturbances in the tracking proce dure. All recursive filters are suitable for real-time implementation. (C) 1997 Academic Press Limited.