International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999 
191 
A LOCAL CORRELATION APPROACH FOR THE FUSION OF REMOTE SENSING DATA WITH DIFFERENT 
SPATIAL RESOLUTIONS IN FORESTRY APPLICATIONS 
J. Hill, C. Diemer, O. Stover, Th. Udelhoven 
Remote Sensing Department, University of Trier, Behringstrasse, Trier, D-54286, Germany 
hill@fews05.uni-trier.de, c.diemer@netcologne.de 
KEYWORDS: Data Fusion, Resolution Enhancement, Local Correlation Modelling, Remote Sensing, Forestry. 
ABSTRACT 
Until now, satellite data are only of limited use to Mid-European forest management. A major limitation is the low spatial resolution 
of the commonly available satellite sensors. In this paper, we present a specific data fusion approach (local correlation modelling) 
which can be used to produce multispectral images with high spatial resolution based on panchromatic reference channels. Such data 
are provided by operational satellite systems (SPOT, IRS-ID, Landsat 7), and their availability may further increase with the advent 
of new commercial satellite systems. Airborne experimental data were used to assess the quality of the modelling approach discussed 
in this contribution, compared to traditionally used fusion algorithms (e.g. Brovey, IHS, PCA, filter techniques). Our validation 
results indicate that local correlation modelling (LCM) performs in all channels significantly better, because the introduced texture is 
locally adjusted to the conditions of each channel. Local contrast differences between a (degraded) panchromatic band and 
multispectral channels are adaptively modelled into the fusion result, even if the local relation between the datasets exhibits an 
inverse contrast polarity. 
1. INTRODUCTION 
1.1. Satellite Data and the Resolution Dilemma 
Until now, spacebome remote sensing data have only been of 
limited use to Mid-European forest management. A major 
limitation is the low spatial resolution of satellite data compared 
to that of the traditionally used aerial photographs. On the other 
hand, satellite remote sensing provides substantial advantages, 
such as high spectral resolution and synoptic coverage of large 
areas. The digital data format allows direct digital processing of 
images and the integration with GIS thematic layers. Another 
important aspect is that remote sensing data can be understood 
as a physical measurement, such that a quantitative analysis 
becomes possible. 
In view of the growing needs for forest information on one hand 
(forest damage, new forest structures), and the necessity to 
reduce costs for inventories on the other hand, these advantages 
are significant. Accordingly, the demand for satellite data with 
higher spatial resolution has increased considerably. Current 
sensor technology allows the deployment of civil high 
resolution satellite sensors, but we find that an increase in 
spatial resolution also leads to an exponential increase in data 
quantity (which becomes particularly important when 
multispectral data should be collected). Since the amount of 
data collected by a sensor has to be balanced against the state- 
of-the-art capacity in transmission rates, archiving and 
processing capabilities (Colvocoresses, 1977), this leads to the 
following dilemma: because of limited data volumes, an 
increase in spatial resolution must be compensated by a 
decrease in other data sensitive parameters, e.g. swath width, 
spectral and radiometric resolution, observation and data 
transmission duration. The reduced swath width also leads to a 
decrease in temporal resolution (which might eventually be 
reduced by multiple and tiltable systems). Summarising, 
improving a satellite sensor’s spatial resolution may only be 
Sensor 
Spatial resolution (m) 
Swath width (km) 
Temporal resolution (days) 
max min (across-track) 
pan 
blue 
green 
red 
NIR SWIR 
TIR 
pan 
multispectral 
TM 7 
15 
30 
30 
30 
30 
30 
120 
185 
185 
16 
- 
SPOT 4 
10 
- 
20 
20 
20 
20 
- 
60 
60 
26 
1-5 
1RS-ID 
5.8 
- 
23 
23 
23 
70 
- 
70 
142 
24 
5 
MOMS 2P* 
6 
18 
18 
18 
18 
- 
- 
36/50 
58/105 
- 
- 
Orb View-3 
1 
4 
4 
4 
4 
- 
- 
8 
8 
16 
<3 
Quick Bird-1 
1 
4 
4 
4 
4 
- 
- 
22 
22 
20 
1-5 
IKONOS-2 
0.8 
3.3 
3.3 
3.3 
3.3 
- 
- 
11 
11 
14 
1-3 
* Priroda Mission. Data for pan refer to the nadir panchromatic channel only. 
Table 1. Characteristics of selected spacebome sensors (modified after Fritz, 1996 ; Aplin et al., 1997).