MULTIVIEW GEOMETRY: PROFILES AND SELF-CALIBRATION
P.R.D.S. Mendonça
2001
An important goal of computer vision is the simultaneous recovery of camera motion and scene structure from multiple views. Frequently, the solution of this problem demands the estimation of the epipolar geometry of the sequence of images, encoded in fundamental matrices. These can be satisfactorily obtained through a number of methods when image features corresponding to the same objects in space --- such as points, lines, texture etc. --- can be easily matched. However, when the scene is comprised of smooth, textureless surfaces, the determination of such correspondences is a difficult problem.
In this situation the most prominent features of the objects being viewed are the profiles or apparent contours. This dissertation develops an efficient technique to estimate the epipolar geometry from profiles in the important case of circular motion. In contrast to previous methods, the solution proposed here can be used even for surfaces with simple geometry. It also employs a sequential approach, obtaining the independent components of the epipolar geometry step-by- step. Finally, it does not demand the solution of any large scale optimisation problem. Once the epipolar geometry is estimated, the projective structure of the scene can be determined. To update this projective reconstruction to an Euclidean one, the camera must be calibrated. In this dissertation a novel self-calibration technique is introduced, based on obtaining the intrinsic parameters that update the fundamental matrices to essential matrices. This simple approach provides an algorithm for the linear computation of the varying focal lengths of the cameras, plus a nonlinear method that can refine the initial solution and also obtain other intrinsic parameters.
If you have difficulty viewing files that end '.gz',
which are gzip compressed, then you may be able to find
tools to uncompress them at the gzip
web site.
If you have difficulty viewing files that are in PostScript, (ending
'.ps' or '.ps.gz'), then you may be able to
find tools to view them at
the gsview
web site.
We have attempted to provide automatically generated PDF copies of documents for which only PostScript versions have previously been available. These are clearly marked in the database - due to the nature of the automatic conversion process, they are likely to be badly aliased when viewed at default resolution on screen by acroread.