Registration via direct methods: a statistical approach

The "direct methods" achieve global image registration without explicit knowledge of feature correspondences. We employ the motion gradient constraint as the relation between the motion parameters and the measured image gradients. While this relation appears as a linear system of equations, for any motion model (other than a translation) we show that the underlying noise process is data-dependent, i.e., heteroscedastic, a fact which must be taken into account in the parameter estimation process. The improvement obtained using the adequate procedure is confirmed for the 2D rigid motion model through comparison with the traditional total least square approach.

[1]  P. Anandan,et al.  Hierarchical Model-Based Motion Estimation , 1992, ECCV.

[2]  Richard Szeliski,et al.  Direct methods for visual scene reconstruction , 1995, Proceedings IEEE Workshop on Representation of Visual Scenes (In Conjunction with ICCV'95).

[3]  P. Anandan,et al.  About Direct Methods , 1999, Workshop on Vision Algorithms.

[4]  Peter Meer,et al.  A general method for Errors-in-Variables problems in computer vision , 2000, Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).

[5]  Michal Irani,et al.  Video indexing based on mosaic representations , 1998, Proc. IEEE.

[6]  Andrew Zisserman,et al.  Feature Based Methods for Structure and Motion Estimation , 1999, Workshop on Vision Algorithms.

[7]  P. Anandan,et al.  Direct Recovery of Planar-Parallax from Multiple Frames , 1999, IEEE Trans. Pattern Anal. Mach. Intell..

[8]  Takeo Kanade,et al.  An Iterative Image Registration Technique with an Application to Stereo Vision , 1981, IJCAI.