Chapter 2: Registration

Title: Automatic registration, mosaicking, and georeferencing of CASI images over large areas

Author(s): Claude Cariou, Kacem Chehdi

Address: IETR/TSI2M – ENSSAT/University of Rennes 1, 6, rue de Kerampont, BP 80518, 22305 Lannion, France | IETR/TSI2M – ENSSAT/University of Rennes 1, 6, rue de Kerampont, BP 80518, 22305 Lannion, France

Reference: Atlantic Europe Conference on Remote Imaging and Spectroscopy pp. 43 - 48

Abstract/Summary: We provide a general methodology for the automatic registration, mosaicking and georeferencing of multiple large multispectral images issued from an airborne imaging scanner, namely the CASI sensor. This methodology is based upon the joint dense registration of flight stripes onto a georeferenced aerial digitised photograph. It uses the mutual information criterion to align both images. This criterion is chosen because it allows the use of images sources with different pattern contents; only the general pattern shapes are taken into account. Another key point of the proposed methodology is to align images in a progressive scheme, from coarse to fine alignment. For this, two steps are set up. In the first one, we try to compensate for deformations due to flight attitude parameters - which are not or imprecisely taken into account in the post-acquisition processing software - by using an affine transformation model. The second step considers a continuous dense displacement map which compensates for remaining misalignments mainly due to the local topography. Mosaicking issues, involving the registration and mixing of multispectral data, as well as georeferencing issues are also considered. We show the efficiency of this methodology for the mosaicking of 6 flight stripes covering 14 km2 at 2 metres resolution.

Order a copy of this article