ve. 
ind 
in 
)99 
ion 
nal 
)ue 
que 
ga 
and 
eo- 
ical 
atic 
] to 
] of 
via 
ates 
aro- 
imi- 
Munro, Duncan 
The initial phase of geometric registration poses some difficulties for images acquired under different sensor-target 
geometry. As geometric distortions arising from terrain relief are dependent upon observation angle the precise registra- 
tion of images acquired with different viewing angles requires the implementation of corrections for relief displacement. 
Consequently, the advent of the ETM- sensor, which acquires panchromatic and multi-spectral data simultaneously (and 
thus with the same senor-target geometry), eliminates a significant amount of processing effort. Whilst a great deal of 
research has been undertaken to develop image fusion products a much smaller effort has been made to assess the value 
of the products within image analysis. Thus the present study focuses on the results from qualitative and quantitative 
comparisons of image fusion products as a tool for information extraction rather than on their production. 
The specific technique used for the generation of image fusion products in this study is referred to within the ERDAS™ 
software package as the Multiplicative Combination. This technique multiplies the values in the multi-spectral data by 
the values of the panchromatic data which results in an increase in the influence of the panchromatic data in the output 
values when compared with techniques based upon ratioing or principal components analysis. This results in a greater 
emphasis on the infrastructure components in the output image which enabled the qualitative comparison of image fusion 
products to be made at a range of scales. 
23. Qualitative Analysis 
The interpretation of LANDSAT TM and ETM+ images and image fusion products for qualitative comparison focused 
on two subject areas. First, the images were compared as tools for interpreting the geology of the study area. Second, the 
images were compared as tools for presenting maps at different scales. The specific interpretive approach for each com- 
parative topic is described below. Inevitably the individual experience of the image interpreter plays an important role in 
the qualitative analysis. No attempt had been made to introduce measures of completeness or accuracy to the qualitative 
analysis by using a series of image analysts to perform the work. 
Common to both topics was the use of colour composites from the image data sets. Colour composites may be used to 
display multi-spectral imagery directly from the DN values within each band and via manipulation of the original values. 
A commonly implemented manipulation produces natural colour composites via a transformation of values in the red, 
green and near infrared bands of the TM and ETM+ sensors. One advantage of the natural colour composite is that the 
imagery shows a familiar appearance to the user and as such is commonly used as a backdrop to thematic maps of land 
cover produced from multi-spectral images. 
2.3.1. Geology 
The availability of geological maps for the study area allowed a comparison to be made between the LANDSAT-5 and 
LANDSAT-7 imagery for identifying lithological and structural elements. This comparison is qualitative and is based 
upon the visual interpretation of the data sets as false colour composites and of the products of image fusion processing 
that was designed to take advantage of the higher resolution panchromatic data (LANDSAT ETM+ Band 8 and SPOT 
PAN images) and the spectral information content of the multi-spectral data. The appearance of the imagery indicates 
that the atmospheric conditions were different for the acquisition of LANDSAT TM and ETM+ data but the objective of 
the comparison between data sets was met without recourse to extensive processing. 
Attention was focussed on the Ordovician-Silurian (500 - 395 m.a.) segment of the stratigraphic column which com- 
prises a series of tightly folded slates, sandstones and quartzites that trend SE-NW and show a strong geo-botanical asso- 
ciation in terms of the amount of cover associated with individual units that permits lithological discrimination within the 
multi-spectral data. This geo-botanical association allows a better discrimination between lithological units to be made in 
the multi-spectral data than with the panchromatic data even thought the latter have a smaller pixel size (10 m for the 
SPOT PAN and 15 m for the LANDSAT ETM+ Band 8 data). The intrusive rocks within the region comprising granites, 
grano-diorites and migmatities show a marked textural pattern that, combined with the cross-cutting nature of their strati- 
graphic boundaries, permits their identification in the multi-spectral data. In the remainder of the stratigraphic column 
the similarity of the units and the extensive agricultural land-use limit the detail of geological interpretation. In addition 
the drainage pattern development shows no association with underlying rocks and as such limited the use of one of the 
traditional means for separating lithological units. Panchromatic imagery from LANDSAT-7 and SPOT were also com- 
pared as tools for identifying the fault lines identified in the geological map base. 
2.3.2. Feature Analysis and Scale Comparison 
One of the major potential improvements provided by the ETM+ in comparison to the TM is the 15 m pixel size panchro- 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 933