By Topic

Total least squares fitting of two point sets in m-D

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
J. A. Ramos ; Dept. of Electr. Eng., Purdue Univ., Indianapolis, IN, USA ; E. I. Verriest

The problem of estimating the motion and orientation parameters of a rigid object from two m-D point set patterns is of significant importance in medical imaging, computer assisted surgery, mobile robot navigation, computer vision, and fingerprint matching. Several least squares algorithms which make use of the singular value decomposition (SVD) have appeared in the literature. These algorithms consider the noise as coming from one image only, when in fact both images are corrupted by noise. In addition, these algorithms may suffer from roundoff accumulation errors due to a SVD of a matrix product between two noise corrupted matrices. This motivates the use of total least squares, where both data sets are treated as noisy. The formulation also avoids computing a SVD of a product of two noise corrupted matrices. This formulation is also convenient for an online implementation. In this paper we treat the image registration problem from a mixed least squares-total least squares point of view. The advantages of such an approach are various: (1) there is no matrix product in the singular value decomposition formulation, (2) the algorithm is given as a nonsingular matrix transformation, T, and (3) it can be easily implemented recursively. Contrary to previous least squares algorithms, the new algorithm takes advantage of the noise structure in the data. That is, we assume there is noise present in both data patterns, as well as a noise-free column for the translation vector. Also, the proposed algorithm computes all the parameters at once, without averaging the data

Published in:

Decision and Control, 1997., Proceedings of the 36th IEEE Conference on  (Volume:5 )

Date of Conference:

10-12 Dec 1997