PROCESS LINE FOR GEOMETRICAL IMAGE CORRECTION OF DISRUPTIVE 
MICROVIBRATIONS 
F. de Lussy 3, *, D. Greslou 3 , L. Gross-Colzy b 
3 CNES SI/QI (Centre National d’Etudes Spatiales) Toulouse France 
francoise.delussy@cnes.fr daniel.greslou@cnes.fr 
b Capgemini South, Space Unit - BP 53655- 31036 Toulouse Cedex 1, France 
lydwine.grosscolzy@capgemini.com 
Commission VI, WG VI/4 
KEY WORDS: Image processing, Correlation, Integration, Rectification 
ABSTRACT: 
Since the beginning of earth observation satellites the dynamic disturbances have a strong impact on satellite design. During 
satellite conception and realisation, they are closely analysed, often with alarmist budgets. ... But so far most satellites have such a 
good stability that these disturbances are even difficult to characterised during in-flight commissioning.. However the recent needs 
to design smaller satellites, more compact and with higher sampling resolution make these dynamic perturbations more probable and 
critical. The correction of these disruptive vibrations thus becomes an important issue, and induces the establishment of a ground 
processing to improve the retrieved satellite attitude. . In this paper, we present a generic algorithm (an integration method) which 
allows us to estimate the disruptive signal from several differential observations derived from imagery. Applied to the PHR system 
in order to secure the budget of geometric stability, this innovative processing gives very accurate results (maximum error is 10% of 
a pixel). . 
1. INTRODUCTION 
Since the first in-flight commissioning of push broom earth 
observation satellite, the inter-retina images spatial matching 
has classically been used to calibrate the.image static models, 
and more precisely the images viewing directions. 
But in fact, because of the push broom system properties, the 
spatial matching which is done between two different times of 
acquisition of the same landscape line by the retina couple, 
produces a measurement of the temporal dynamic geometrical 
behaviour of the spatial images. This differential geometrical 
profile can be related to residues of attitude out of the dynamic 
model, and hence is of particular interest. Indeed, the on-board 
system AOCS estimates satellite attitude with cut-off 
frequencies in the range of 8 to 16Hz and thus cannot retrieve 
the exact image attitude which is subject to the effects of higher 
frequency vibrations on mirrors, focal plane etc... 
The differential attitude residues estimated from the inter-retina 
images spatial matching are therefore measured in terms of 
specific image quality needs. By increasing the field of in-flight 
commissioning, and processing images of each product, it 
becomes possible to use this differential data residues of 
attitude to actually correct the images of disruptive 
microvibrations. The integration of these various differentials 
would allow us to restore the absolute disruptive signal. 
But, as the differential measurements are coming from imagery, 
we have to keep in mind that we are restricted by the pixels 
time sampling, the record length along the satellite track, and 
the high measurement noise coming from the spatial matching. 
The a posteriori correction will also not be effective on raw 
image quality budgets like local coherence (spatial local 
sampling) and will not allow us to correct the geometrical 
effects on Modulation Transfer Function (row and column 
desynchronization that affect the Basic radiometric information). 
This line of sight attitude amelioration can be done under 
certain assumptions concerning the characteristics of the 
various retinas mounted in the focal plane, the sampling of 
images, and the disturbance signals that we want to correct. In 
fact, attitude differentials measurements are possible if inter 
retina arrays are almost parallel and if the images produced by 
these retinas correlate sufficiently. In addition, the frequency 
range is reduced by the conditions of measurement of the inter 
retinas images spatial matching : the disruptive signal is 
convoluted by a moving-average filter of size the number of 
lines of the picture window used for correlation, and constraints 
appear coming from the time lapse between the different retinas 
couples. Finally the integration process may differ from a 
viewing instrument to another, because the disruptive signal 
depends on the mechanical vibration source, equipment which 
may amplify these sources and therefore their signals signatures 
(frequencies law, magnitude law, the phases law). 
We present an overview of the main concept of dynamic 
correction, starting with an application on SP0T5, and followed 
by a focus on the new ground processing line applied to image 
products for the restitution of microvibrations. This ground 
processing line has been prototyped during the years 2006/2007 
Corresponding author