“3-D Estimation Using a Sequence of Noisy Stereo Images Part II: Uniqueness Results”
by Gem-Sun Young and Rama Chellappa
We discuss the uniqueness of motion parameter estimates in the kinematic model based approach presented in a companion report [3]. The kinematic model is designed for the estimation of 3-D motion parameters from a sequence of noisy stereo images. A constructive proof of uniqueness is established by separating the rotational parameters from the translational parameters. We show that with uniform sampling, three noncollinear feature points in two consecutive binicular image pairs and one feature point each in three binocular image pairs contain all the spatial and temporal information of the translational motion with constant acceleration and rotational motion with constant angular velocity. Some related issues such as sampling of image sequences and ambiguity in motion are also analyzed.