desire for automation of repetitive routine interpretations. Geologists/
planetologists study essentially static phenomena, where the need for auto-
mation is less obvious and where desired results may be more difficult to
achieve.
Pattern recognition has a longer tradition. It began with research on
automatically recognizing symbols using non-natural . images, i.e. two-
dimensional formats for general data presentation (Andrews and Hunt, 1977).
With the advent of digital natural images it was logical to subject them
to pattern recognition techniques. Today the terms pattern recognition
and ‘image processing sometime, are used synonymously when in fact, how-
ever, they denote technologies that overlap only partly.
Digital image processing methods may be subdivided in various ways. Goetz
et al. (1975) for example use the following grouping of methods: (a) recti-
fication - elimination of systematic geometric and radiometric errors, (b)
cosmetics - elimination of random noise and defects, (c) analysis - data
extraction,(d) display. A similar grouping of processing techniques is used
by Pratt (1978): (a) restoration, (b) enhancement, (c) analysis, (d) image
coding. In this context classification with multispectral scan (MSS) ima-
ges is one of many analysis techniques.
To many & photo-interpreter, however, image processing techniques fall
in one of two groups: pre-processing and automatic data analysis (classi-
fication or feature extraction). This view is shared for exampleby Holder-
man et al. (1976) and Anuta (1977).
The paper presents a review of techniques to pre-process digital images
for subsequent photo-(image) interpretation, using recent experiences and cur-
rent results. In our discussion of pre-processing methods we separately
address single panchromatic images and multiple images (MSS) as well as
image synergisms. This latter concept denotes techniques of compositing
images from different sources (e.g. radar/Landsat etc.).
Digital image processing is a complex technology that has been developed
in an environment in which the actual users are not always involved. There-
fore there exists a user/technologists interface problem that we discuss
in & concluding section.
2. WHY DIGITAL PRE-PROCESSING ?
The need for digital techniques in automatic analysis (information
extraction, decision making) from images is obvious. Far less obvious is
this need when it concerns the preparation for subsequent traditional photo-
interpretation.
Aerial photography has been electronically dodged for a long time to
improve contrasts. It seems obvious that telemetered images that present
themselves in a digital format undergo digital contrast manipulations.
Density requantisation (Histogram equalisation®) and contrast stretching
are indeed the most common and basic techniques of pre-processing (Goetz
et al., 1975).
"Histogram equalisation" in the U.S.A. may be understood to relate to
the correction of LANDSAT six —line striping.
cr H FE*cFO z zz H Hi
ph
ne
We
ve
