The development of methods to detect slowly progressing diseases is often hampered by the time-consuming acquisition of a sufficiently large data set. In this paper, a method is presented to model the change in images acquired by scanning laser polarimetry, for the detection of glaucomatous progression. The model is based on image series of 23 healthy eyes and incorporates colored noise, incomplete cornea compensation and masking by the retinal blood vessels. Additionally, two methods for detecting progression, taking either one or two follow-up visits into account, are discussed and tested on these simulated images. Both methods are based on Student's t-tests, morphological operations and anisotropic filtering. The images simulated by the model are visually pleasing, show corresponding statistical properties to the real images and are used to optimize the detection methods. The results show that detecting progression based on two follow-up visits greatly improves the sensitivity without adversely affecting the specificity.

In this paper we describe a new system for the reconstruction of geometric map elements as developed in the Dutch TopSpin-PNEM project 'Knowledge-based conversion of utility maps'. In the knowledge base map conversion process vectorized map elements coming from an image interpretation system must be reconstructed in a geometrically accurate topographic base map. This is done using a map correspondence between the utility map and the base map based on mutual topography and relational information supplied by a map interpretation module. The map interpretation module uses geometrical inference to deduce the relational information, while applying a novel conflict resolution strategy. The map elements are accurately reconstructed, relationally and geometrically, using the map correspondence, the relations and several object-specific functions.

In this paper an existing ridge detector based on the Hessian tensor and a new one based on the structure tensor are evaluated for use in portal/simulator imaging. Tensor smoothing is introduced as a scaling parameter. By changing this scale (as opposed to the image scale) the detectors can be tuned to suppress spurious ridges without moving them. Desirable ridge detector properties for portal/simulator imaging are described and the ridge definitions are tested accordingly. Experiments show that our new ridge detector outperformed the existing one in simulator/portal image matching.

The Euclidean skeleton is essential for general shape representation. This paper provides an efficient method to extract a well-connected Euclidean skeleton by a neighbor bisector decision (NBD) rule on a vector distance map. The shortest vector which generates a pixel's distance is stored when calculating the distance map. A skeletal pixel is extracted by checking the vectors of the pixel and its 8 neighbors. This method succeeds in generating a well-connected Euclidean skeleton without any linking algorithm. A theoretical analysis and many experiments with images of different sizes also shows the NBD rule works excellent. The average complexity of the method with the NBD rule algorithm and the vector distance transform algorithm is linear in the number of the pixels.

An algorithm is presented that allows one to perform skeletonization of large maps with much lower memory requirements than with the straightforward approach. The maps are divided into overlapping tiles, which are skeletonized separately, using a Euclidean distance transform. The amount of overlap is controlled by the maximum expected width of any map component and the maximum size of what is considered as a small component. Next, the skeleton parts are connected again at the middle of the overlap zones. Some examples are given for efficient memory utilization in tiling an A0 size map into a predefined number of tiles or into tiles of a predefined (square) size. The algorithm is also suited for a parallel implementation of skeletonization.

A general contour-oriented shape dissimilarity measure is shown to give unreliable answers due to the arc-length parametrization. Local differences in curve-length due to biological variation cause mismatches in the contour-representations. In many applications all objects to be compared have a basic model in common, which can often be found by the use of landmarks. In the application shown here, which was a comparison of flowers of potato-species (Solanum spp), the basic model was characterized by the five petal tips, which should be at equal intervals. By restoring the basic model for each contour before calculating the dissimilarity measures, the errors of the arc-length parametrization were considerably reduced. This resulted in more reliable and faster dissimilarity measures for the application of potato flowers.

In automatic line drawing interpretation (e.g. map interpretation), one of the problems encountered is the finite size of scanners, or that scanners of the required size are not available. Often, one large scan is necessary for the interpretation process, instead of several smaller ones generated by the usual scanning in parts. This paper describes a method for automatically compiling mosaics from separately scanned line drawings. A mosaic is a collection of separately obtained images which are combined to form one larger image. The method is based on vectorization of the line drawings, which is used to select the control points for a geometric transformation automatically. It is not necessary to specify the overlap area between the line drawings. The resulting system is evaluated using large scale maps. Experiments with different overlaps between the line drawings were done. Results are good: the algorithm succeeds in finding accurate parameters for the transformation.>

A base line system for automatic interpretation of Dutch cadastral maps is discussed. Knowledge of the rules for drawing these maps is incorporated in the system. Also, a method for evaluating this system is presented. This method consists of different parts which subsequently evaluate the vectorization, the performance on finding parcels, and the performance on finding parcel numbers. Parts of this evaluation method may also be applicable for other line drawing interpretation systems.>

The probability density function is computed of the radius of all circular discs that can be digitized into a distinct small number of pixels (>

The speed of skeletonizing algorithms is improved by applying them on contour pixels only. Further improvement can be gained by using chain-coded contours. If the deletion of a contour pixel is allowed, according to a simple connectivity test, the corresponding pair of chain codes is replaced by a predefined test of other chain codes. The resulting skeleton does not require cleaning or pruning, but the number of branches can be easily controlled by modifying the replacement table. Since the algorithm is sequential, the starting pixel on the contour requires special attention to avoid spurious branches.>