Home Research Teaching Personal

  Main   Mathematical Models in Computer Vision: The Handbook, Springer (2005)

  • Image Alignment and Stitching, Richard Szeliski
ABSTRACT: Stitching multiple images together to create beautiful high-resolution panoramas is one of the most popular consumer applications of image registration and blending. In this chapter, I review the motion models (geometric transformations) that underlie panoramic image stitching, discuss direct intensity-based and feature-based registration algorithms, and present global and local alignment techniques needed to establish high-accuracy correspondences between overlapping images. I then discuss various compositing options, including multi-band and gradient-domain blending, as well as techniques for removing blur and ghosted images. The resulting techniques can be used to create high-quality panoramas for static or interactive viewing.

Last Update: December20th, 2004, you can mail your comments to: nikos.paragios@computer.org