n 
A cigital image is the digital description and ordered 
Storage of the position and the density values of finitely 
small image elements (pixels). The direct digital recording 
of sensor information can be distinguished from the later 
digitizing of imagery previously recorded by analogue espe- 
cially photographic means. 
Digital image processing is in first instance computational 
manipulation of aray-scale density values (apart from correc- 
tion of geometrical data), for the purpose of general or 
special improvement of image quality and of interpretability. 
Such image enhancement is concerned with elimination of par- 
tial sensor or transfer disturbances, with the filtering of 
noise, the modification of average values and range of den- 
sities or of contrast. Genuinely improved images can be 
obtained and photographically redisplayed. 
Comparison of two or more images with space- or time-separa- 
tion allows monitoring of differences or alterations of 
contents. Also digital correlation of image details can be 
used for measuring parallaxes or for geometrical matching 
of different images. 
The second main aim of digital image processing is pattern 
recognition, or rather more specific object classification 
for eventual automatic image interpretation. Object classi- 
fication attempts to identify types of objects by assessing 
and analyzing density values of multiple image elements from 
different bands of wavelengths of multispectral imagery. 
There are, however, considerable difficulties. The spectral 
features of various classes of objects (such as vegetation, 
crops etc.) do not differ very much, range of dispersion is 
large, and disturbing effects (sun angle, shadows) have to 
be eliminated. 
Digital image processing is developed and investigated with 
great efforts in a surprisingly large number of research 
institutions, recently also outside the USA. The fully 
automatic photo-interpretation for topographic mapping re- 
mains an utopian aim, for the time being. Nevertheless 
successful classification of vegetation and land use has 
been achieved. Also highly successful applications are re- 
ported from special fields, such as glaciology, astronomy, 
medicine, nuclear physics. The procedures for image enhan- 
cement, for change detection, and for manipulating multi- 
stage imagery have reached a high level of development with 
various practical applications. Most spectacular examples 
are processings of images from Mars and Mercury. 
Conventional photogrammetry has also drawn some benefits from 
the development of image processing, especially with regard 
to digital image correlation for the purpose of automated 
orthophotography or for automatic point transfer.