
5 Calibrating Raster Images
Calibrating (also known as “rubbersheeting”) is designed to eliminate arbitrary (both
linear and non-linear) distortions in monochrome, grayscale, and color raster
images: scanned graphic documents, geodetic plans, maps in raster format, etc.
To carry out calibration original documents must include points with known
coordinates. These points can have a different position in the scanned image,
caused by the deformation in original materials or scanning errors. By applying
calibration, raster images are transformed so that the current coordinates of such
points coincide with their known values.
Calibration can be applied to a selection set consisting of image fragments and also
to a clipped area on the image. If several images have been selected, then the
command will be applied to those that are visible and located on unlocked layers. If
an image is not selected, then this command processes all visible images located
on unlocked layers. The calibration procedure needs to be applied before running
the procedures of layering and vectorization. If there is non-linear distortion in the
original image, then after vectorization a vector drawing is obtained, which is
impossible to correct. To obtain a correct vector drawing, eliminate distortion before
vectorization..
Terms
The calibration procedure transforms a raster image in such a way that the given
set of image points moves to another set of points with pre-defined coordinates.
The number of points and their locations are arbitrary.
The calibrating transformation is defined by the set of
calibration pairs
and the
calibration method used
.
Calibration Pairs
To prepare calibration, specify vectors of point movements. This can be done by
specifying a set of calibration pairs. Each pair represents two locations of the point -
its current location on the image (
measured point
), and its required location (
real
point
).
Calibration Methods
A
calibration method
is a parametric transformation function used for calibrating.
The program offers a number of calibration methods: affine, bilinear, polynomial,
surface splines, etc.
With some sets of calibration pairs and some calibration method the program
cannot make a transformation of the given type that where all measured points
move to appropriate real points. This produces calibration deviation distances
between the points obtained as a result of transformation and their appropriate real
points. When choosing a calibration method, use the minimizing of the square error
value for all calibration pairs as criteria.