Autodesk 05720-017408-9641 User Guide - Page 54

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.