There has been proposed the image compression method based on two-dimensional orthogonal expansion by use of marginal eigenvectors of the covariance matrix of an image. A method for image registration based on the two-dimensional orthogonal expansion by marginal eigenvectors for a set of training images. The method is very efficient by use of the vectors with smaller dimensions rather than the number of pixels. In this paper, we realize a rotation-invariant registration method based on the method, and then verify the effectiveness of the proposed method through fundamental experiments with real images. Furthermore, two algorithms for fast computation are proposed. Experiments for verification of effectiveness are planned.
展开▼