344
contains some concepts and tools for theoretical quality assessment.
The contents are arranged according to the sequence of data flow
indicated in figure 1. Emphasis is placed on the most crucial and
complex problems.
II. INFLUENCING FACTORS AND SOURCES OF DISTURBANCES
Influencing factors may have negative and/or positive effect on
performance and reliability. Disturbances originate in approximate
algorithms, physical processes, incrementation of data (spatial,
intensity), filtering (and interpolation), thresholding, and even-
tually in rounding-off numerical data. The last, however, can be
made insignificant and will therefore not be considered further.
II.l Scanning-digitizing
The factors and sources involved can be differentiated into the
geometric and pictorial (intensity). Geometric factors are the spa-
tial step (increment) between adjacent pixels, pixel size, and use
(or not) of the geometry of the photogrammetric model (e.g., epipo-
lar geometry). Geometric disturbances originate in relative orienta-
tion (if model gemoetry is used), in sensor positioning, and in the
spatial lay-out (i.e., step size and phase) of pixels.
Pictorial factors concern illumination intensity, sensor acuity,
size of sensing area, integration time at sampling, and intensity
increment size. Disturbances originate in uneven illumination, vary-
ing responsivity of sensors, noise, and in incrementation of
intensity. Á
II.2 Pre-processing
If properly applied pre-processing upgrades raw image data. Never-
theless, the operations involved [4| cause some losses in both geo-
metric and pictorial domains. These losses can be suppressed and
some gains achieved by merging different input data sets, excluding
anomalous regions, correcting data, analysing data and extracting
distinct features, and by synthesising new image data. Significant
losses, however, can be attributed to resampling, image transorma-
tions (intensity), and filtering and thresholding (for compression).
Data segmentation and structuring can improve the quality of image
matching at the expense of reduced time-efficiency. This also holds
for integration of external information (to be used at image match-
ing and/or post-processing). Inadequate data segmentation and struc-
turing, and improper use of external information may cause substan-
tial losses in both terms of quality and time-efficiency.
II.3 Image matching *
The most essential factors concern the strategy, algorithms and
techniques of matching [4]. These are strongly interdependent and
have both positive and negative effects on performance. More sophis-
ticated procedures tend to increase accuracy and reliability of
matching, but are less time-efficient, and vice versa. In off-line
systems, however, time-efficiency is less important.
Disturbances can be attributed to image data (i.e., distortions) and
345
(UM
