We propose an efficient algorithm for image alignment. The key step is to estimate the inverse of the incremental update to the warp. The incremental update is then inverted and composed with the current estimate to form the new estimate for the next iteration. The set of warps must therefore form a group. The algorithm is efficient because estimating the inverse of the update reverses the role of the two images. Much of the computation then only needs to be performed once, rather than once for each iteration. Keywords: Backwards incremental image alignment and registration, efficiency. 1 Introduction Image registration or alignment consists of moving, and possibly deforming, a template to minimize the difference between the template and an image. Some of the applications of alignment include optical flow [Lucas and Kanade, 1981], tracking [Black and Jepson, 1998, Hager and Belhumeur, 1998, Cascia et al., 2000], parametric (or layered) motion estimation [Bergen et al., 1992], mosaic...
CITATION STYLE
Baker, S., Dellaert, F., & Matthews, I. (2001). Aligning Images Incrementally Backwards. Pittsburgh; PA. Retrieved from http://www.ri.cmu.edu/pubs/pub_3491.html
Mendeley helps you to discover research relevant for your work.