is known as the World Reference System (WRS), and is peculiar to that satellite. Other satellites, 
such as SPOT, MOS-1 and ERS-1 have a similar reference system. 
In contrast to the satellite-specific coordinate systems, geocoded data is aligned with the 
geographic grid as shown in Figure 5. Geocoded data sets are: 
1. Corrected for all source-dependent errors 
2. Geometrically transformed into the desired map projection. 
3. Rotated to align with the map axes. 
4. Resampled to a standard square pixel size. 2 
5. Usually sized in units of an integral number of standard map sheets. 
Data processed in this way can 
be easily manipulated along with 
other geographic data, and 
used in hardcopy form in exactly 
the same way that a map is 
used. In addition, the pixels are 
oriented in rows which run east- 
west, and columns which run 
north-south, with each pixel tied 
to a geographic location which is 
predetermined by the particular 
map reference system chosen. 
In other words, we predefine a 
fixed set of pixel locations on 
the ground, and as each 
satellite acquisition is made, the 
resulting measurements of the 
radiance distribution are 
transformed into a set of 
numbers which represents the 
same radiance distribution, but 
expressed in terms of the 
radiance coming from each of 
the predefined pixels. This 
involves a process known as resampling, in which the original data (in the original coordinate 
system) is effectively filtered to recreate the original radiance distribution function, and then 
resampled into the new coordinate system. The degree to which geometric resampling of digital 
image data pollutes the original radiance measurement has been a hotly debated issue for many 
years. Any digital geometric manipulation of the data involves a resampling operation. Before 
considering the use of resampling techniques, it is important to understand that all current 
spacecraft data is undersampled in at least a part of the image. This implies that in those parts of 
the image, the data is aliassed, and is therefore in error. Anv resampling operation will cause that 
part of the image to be transformed with an error depending on the amount of aliassing present. 
Since the utility of the data sets is greatly improved by geometric rectification, and successive 
resampling operations compound the distortions introduced by a single resampling operation, it is 
important that resampling be done only once, and that it be done with the maximum achievable 
fidelity. By resampling to a standard geographic grid, the data is converted into a form which can 
be combined with other data sets easily without the necessity of further resampling. 
Figure 5 A Geocoded Data Set and 
Relationship to the Satellite track 
2 The standard pixel sizes chosen are integral multiples of one another. This allows superposition of images 
of different pixel resolutions, and the cross-interpretation of radiance/reflectance readings from such 
images since the pixels nest precisely. A common set of standard pixel sizes are: 50 meters (Landsat 
MSS), 25 meters (Landsat TM and Seasat SAR),12.5 meters (SPOT MLA), and 6.25 meters (SPOT PLA). 
9