An Automated Image Registration System For SPOT Data 
D Boardman, | J Dowman, A Chamberlain, D Fritsch, W Newton 
KEYWORDS: Registration, Automation, Satellite, Image, Remote Sensing 
PURPOSE: 
This paper reports the development of a software system which automatically registers one SPOT 
image to another. The work is the second phase of a contract for the Western European Union 
Satellite Centre (WEU). The system is known as the Prototype Automatic Image Registration 
System (PAIRS) and has been carried out by Earth Observation Sciences Ltd, University College 
London, the Institute for Photogrammetry of the Technical University of Stuttgart and the 
University of Oporto. Future phases of the work will incorporate a three dimensional 
transformation and allow the co-registration of imagery from different sensors. 
1. SYSTEM OVERVIEW 
1.1 Rationale of the System 
Image registration is a necessary precursor to 
many remote sensing applications, such as the 
detection of change. Manual selection of the 
large number of tie points required to achieve 
the required accuracy of image registration is 
often a laborious process. The system 
described below takes two panchromatic SPOT 
1A images which have at least 5096 overlap and 
registers the two images. This is achieved by 
the automated determination of tie points. Once 
these tie points are determined a geometric 
transformation is calculated between the two 
images and the second image resampled to co- 
register it with the first image. 
1.2 Process Overview 
Images are co-registered by the automated 
determination of tie points. From a set of tie 
points defining conjugate points in two images 
an n dimensional transformation can be defined. 
This is applied to an image pair, such that the 
second image is resampled to the first image. 
The heart of the system is a set of processes 
which seek to define a set of reliable tie points 
which can be used in the definition of an image 
to image transformation. Possible tie points 
are found on each image of an image pair. 
These sets of points are considered and 
conjugate point pairs determined. A major part 
of the system is thus essentially an image 
matching system. The matching is of point 
features. Future phases on the system will 
attempt to match on the basis of other features 
such as regional features. 
An important aspect of the system is that it is 
automated, or at least as automated as 
possible. The procedure of defining a set of 
good tie points, and hence good geometric 
transformation, is iterative, with a series of 
refining procedures removing tie points for 
which there is no strong evidence. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996