Global Methods for Image Motion Analysis
Abstract
In this thesis, we are concerned with the situation where the optical flow has already been computed and we are interested in determining the egomotion parameters. The problem is rendered difficult because of the following two reasons [4]: (1) The relationship between the three-dimensional motion parameters and the optical flow is nonlinear, and (2) the space of unknowns has a dimensionality of five. In other words, the straightforward methods to solve the problem are non-linear and involve minimization in a five-dimensional space. In developing our algorithms, we exploit the fact that even though the relationship between the parameters and the optical ow is nonlinear, it is actually bilinear (in some parameters). Certain observations about the structure of the equations that describe this relationship enable us to reduce the dimensionality of the search space to two. We also provide approximate algorithms that involve no search. One important feature of all the algorithms is that they are global; that is, they use all the available information in the image optical flow data to compute the parameters of motion.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1992
- Accession Number
- ADA598278
Entities
People
- Venkataraman Sundareswaran
Organizations
- New York University