Klaus D. Tönnies

Line detection and segmentation in historical church registers

For being able to automatically acquire the information recorded in church registers and other historical scriptures, the writing on these documents has to be recognized. This paper describes algorithms for transforming the paper documents into a representation of text apt to be used as input for an automatic text recognizer. The automatic recognition of old handwritten scriptures is difficult for two main reasons. Lines of text in general are not straight and ascenders and descenders of adjacent lines interfere. The algorithms described in this paper provide ways to reconstruct the path of the lines of text using an approach of gradually constructing line segments until a unique line of text is formed. In addition, the single lines are segmented and an output in form of a raster image is provided. The method was applied to church registers. They were written between the 17th and 19th Century. Line segmentation was found to be successful in 97% of all samples.

A New Approach for Model-Based Adaptive Region Growing in Medical Image Analysis

Interaction increases flexibility of segmentation but it leads to undesired behaviour of an algorithm if knowledge being requested is inappropriate. In region growing, this is the case for defining the homogeneity criterion as its specification depends also on image formation properties that are not known to the user. We developed a region growing algorithm that learns its homogeneity criterion automatically from characteristics of the region to be segmented. It produces results that are only little sensitive to the seed point location and it allows a segmentation of individual structures. The method was successfully tested on artificial images and on CT images.

Prior Shape Level Set Segmentation on Multistep Generated Probability Maps of MR Datasets for Fully Automatic Kidney Parenchyma Volumetry

Fully automatic 3-D segmentation techniques for clinical applications or epidemiological studies have proven to be a very challenging task in the domain of medical image analysis. 3-D organ segmentation on magnetic resonance (MR) datasets requires a well-designed segmentation strategy due to imaging artifacts, partial volume effects, and similar tissue properties of adjacent tissues. We developed a 3-D segmentation framework for fully automatic kidney parenchyma volumetry that uses Bayesian concepts for probability map generation. The probability map quality is improved in a multistep refinement approach. An extended prior shape level set segmentation method is then applied on the refined probability maps. The segmentation quality is improved by incorporating an exterior cortex edge alignment technique using cortex probability maps. In contrast to previous approaches, we combine several relevant kidney parenchyma features in a sequence of segmentation techniques for successful parenchyma delineation on native MR datasets. Furthermore, the proposed method is able to recognize and exclude parenchymal cysts from the parenchymal volume. We analyzed four different quality measures showing better results for right parenchymal tissue than for left parenchymal tissue due to an incorporated liver part removal in the segmentation framework. The results show that the outer cortex edge alignment approach successfully improves the quality measures.

A genetic algorithm for automated horizon correlation across faults in seismic images

Finding corresponding seismic horizons which have been separated by a fault is still performed manually in geological interpretation of seismic images. The difficulties of automating this task are due to the small amount of local information typical for those images, resulting in a high degree of interpretation uncertainty. Our approach is based on a model consisting of geological and geometrical knowledge in order to support the low-level image information. Finding the geologically most probable matches of several horizons across a fault is a combinatorial optimization problem, which cannot be solved exhaustively since the number of combinations increases exponentially with the number of horizons. A genetic algorithm (GA) has been chosen as the most appropriate strategy to solve the optimization problem. Our implementation of a GA is adapted to this particular problem by introducing geological knowledge into the solution process. The results verify the suitability of the method and the appropriateness of the parameters chosen for the horizon correlation problem.

Stable dynamic 3D shape models

Shape models used for the segmentation of 3D image data often suffer from high instability of shape. Current approaches to avoid this instability often result in models with high computation times and few possibilities for interaction and modelling. We present a 3D mass-spring model which has been extended by torsion forces and the capability of explicit rotation. These models are stable with respect to shape collapse and contortion. Stability is achieved even if the model is only sparsely connected. This makes the computation efficient enough for real-time interaction. The extended model has been successfully applied to the segmentation of the left ventricle of the human heart in 3D SPECT data.

Focal region-guided feature-based volume rendering

In this paper we advocate the use of a focal region-guided feature-based volume renderer that offers an alternative for visualization of internal structures of volumetric data. We describe this promising technique for communicating the first impression of object shape or contour while at the same time providing detailed information of volumetric data with the use of a lens-like focal region. We designed a method for generating object contours and enhancing volumetric features to depict context information out of the focal region based on gradient volume. In the focal region, we render interesting volume data using the direct volume rendering method. The rendering is guided by the focal region through which we can specify what subset to render. The connection between the structure of the interior and exterior of the focal region gives a better understanding of the spatial relationship. This is to communicate better the existence, form, and location of underlying targets while minimally occluding them.

Distance Based Enhancement for Focal Region Based Volume Rendering

This paper proposes a method to control the object enhancement in context region in focal region based volume rendering. The proposed method uses the distance as a factor to control the volume features in the context region. The main contributions are as follows: To introduce the distance into the rendering pipeline for volume feature enhancement; To demonstrate the implementation of how to use distance in focal region based volume rendering; And to show the important capabilities of distance based enhancement in focal region based volume rendering for 3D data interpretation.

Automatic segmentation of the left ventricle in 3d SPECT data by registration with a dynamic anatomic model

We present a fully automatic 3D segmentation method for the left ventricle (LV) in human myocardial perfusion SPECT data. This model-based approach consists of 3 phases: 1. finding the LV in the dataset, 2. extracting its approximate shape and 3. segmenting its exact contour. Finding of the LV is done by flexible pattern matching, whereas segmentation is achieved by registering an anatomical model to the functional data. This model is a new kind of stable 3D mass spring model using direction-weighted 3D contour sensors. Our approach is much faster than manual segmention, which is standard in this application up to now. By testing it on 41 LV SPECT datasets of mostly pathological data, we could show, that it is very robust and its results are comparable with those made by human experts.

Using an Active Shape structural model for biometric sketch recognition

A deformable shape model called Active Shape Structural Model (ASSM) is used within a biometric framework to define a biometric sketch recognition algorithm. Experimental results show that mainly structural relations rather than statistical features can be used to recognize sketches of different users with high accuracy.

Deformable structural models

A hierarchical framework for the recognition of complex deformable shapes is developed. In extension to traditional approaches an additional layer of control is introduced to guide the local search for subshapes. This is realized by incorporating knowledge about their spatial relationships. A new technique of expectation maps is applied to allow simultaneous shape searches to influence each other. Furthermore, these maps are used to assess spatial coherence among shapes. Thus, the occurrence of well matched shapes at some places in the image may suggest searches for related shapes at other positions. An application to classify species in ant image databases shows promising initial results.