VisMed Research Projects

[sample image] Direction-Driven Shape-Based Interpolation of Volume Data (2001)
We present a novel approach to shape-based interpolation of gray-level volume data. In contrast to the segmentation-based techniques our method directly processes the scalar volume requiring no user interaction. The key idea is to perform the interpolation in the directions given by analysis of the eigensystem of the structure tensor. Our method processes a 256 x 256 slice within a couple of seconds yielding satisfactory results. We give a quantitative and a visual comparison to the linear inter-slice interpolation. Analysis of the results lead us to the conclusion that our technique has a strong potential to compete with well-established shape-based interpolation algorithms.
[sample image] EndoView: A Phantom Study of a tracked virtual bronchoscopy (2001)
Virtual endoscopy can be used for preoperative planning, for training and intraoperatively. Surface rendering displays the inner lumen very well. Volume rendering has to be used if the external structures are of interest. For certain applications, e.g. endoluminal biopsy, it is of great advantage to be able to use both techniques at once. In this work we describe an approach that allows to use these two methods in combination on a low-end standard personal computer. Since image generation is done in a preprocessing step, any high quality volume or polygonal rendering technique can be used and mixed together without any loss in performance at run-time. This work extends a previous image based rendering system for virtual bronchoscopy to include tracking of a rigid or flexible endoscope and finding one's way in the tracheal tree by displaying the endoscope's position in a top-view map of the trachea. Natural landmarks, i.e. bifurcations in the bronchial tree, are used for registration. Properties of the technique are explored on a phantom data set.
[sample image] Fast Visualization of Object Contours by Non-Photorealistic Volume Rendering (2001)
In this paper we present a fast visualization technique for volumetric data, which is based on a recent non-photorealistic rendering technique. Our new approach enables alternative insights into 3D data sets (compared to traditional approaches such as direct volume rendering or iso-surface rendering). Object contours, which usually are characterized by locally high gradient values, are visualized regardless of their density values. Cumbersome tuning of transfer functions, as usually needed for setting up DVR views is avoided. Instead, a small number of parameters is available to adjust the non-photorealistic display. Based on the magnitude of local gradient information as well as on the angle between viewing direction and gradient vector, data values are mapped to visual properties (color, opacity), which then are combined to form the rendered image (MIP is proposed as the default compositing stragtegy here). Due to the fast implementation of this alternative rendering approach, it is possible to interactively investigate the 3D data, and quickly learn about internal structures. Several further extensions of our new approach, such as level lines are also presented in this paper.
[sample image]
[sample image]
Virtual Colon Unfolding (2001)
The majority of virtual endoscopy techniques tries to simulate a real endoscopy. A real endoscopy does not always give the optimal information due to the physical limitations it is subject to. In this paper, we deal with the unfolding of the surface of the colon as a possible visualization technique for diagnosis and polyp detection. A new two-step technique is presented which deals with the problems of double appearance of polyps and nonuniform sampling that other colon unfolding techniques suffer from. In the first step, a distance map from a central path induces nonlinear rays for unambiguous parameterization of the surface. The second step compensates for locally varying distortions of the unfolded surface. A technique similar to magnification fields in information visualization is hereby applied. The technique produces a single view of a complete virtually dissected colon.
[sample image] Virtual Colon Flattening (2000)
We present a new method to visualize virtual endoscopic views. We propose to flatten the organ by the direct projection of the surface onto a set of cylinders. Two sampling strategies are presented and the introduced distortions are studied. A non-photorealistic technique is presented to enhance the perception of the images. Finally, an approximate but real-time endoscopic fly-through is possible by using the data obtained by the projection technique.
[sample image] Interactive Volume Rendering based on a ``Bubble Model'' (2000)
In this paper an interactive volume rendering technique is presented which is based on a novel visualization model. We call the basic method ``bubble model'' since iso-surfaces are rendered as thin semi-transparent membranes similarly to blown soap bubbles. The primary goal is to develop a fast previewing technique for volumetric data which does not require a time consuming transfer function specification to visualize internal structures. Our approach uses a very simple rendering model controlled by only two parameters. We also present an interactive rotation technique which does not rely on any specialized hardware, therefore it can be widely used even on low-end machines. Due to the interactive display, fine tuning is also supported since the modification of the rendering parameters has an immediate visual feedback.
[sample image] Salient Representation of Volume Data (2000)
We introduce a novel approach for identification of objects of interest in volume data. Our approach tries to convey the information contained in two essentially different concepts, the object's boundaries and the narrow solid structures, in an easy and uniform way. The second order derivative operators in directions reaching minimal response are involved for this task. To show the superior performance of our method, we provide a comparison to its main competitor - surface extraction from areas of maximal gradient magnitude. We show that our approach provides the possibility to represent volume data by its subset of a nominal size.
[sample image] Cylindrical Approximation of Tubular Organs for Virtual Endoscopy (2000)
Virtual endoscopy is a promising medical application of volume visualization techniques. A virtual endoscopy system requires high quality and perspective projection rendering, as well as real-time navigation. In this paper the generation of a cylindrical structure for tubular shaped organs (i.e. colon, aorta) is presented. This structure represents an approximation of the real organ. The cylindrical structure will be used to accelerate high quality volume rendering.
[sample image] Exploiting Eigenvalues of the Hessian Matrix for Volume Decimation (2000)
In recent years the Hessian matrix and its eigenvalues became important in pattern recognition. Several algorithms based on the information they provide have been introduced. We recall the relationship between the eigenvalues of Hessian matrix and the 2nd order edge detection filter, show the usefulness of treating them separately and exploit these facts to design a combined threshold operation to generate sparse data sets.
[sample image] Mastering Interactive Surface Rendering for Java-Based Diagnostic Applications (2000)
The display of iso-surfaces in medical data sets is an important visualization technique used by radiologists for the diagnosis of volumetric density data sets. The demands put by radiologists on such a display technique are interactivity, multiple stacked transparent surfaces and cutting planes that allow an interactive clipping of the surfaces. This paper presents a Java based, platform independent implementation of a very fast surface rendering algorithm which combines the advantages of explicit surface representation, splatting, and shear-warp projection to fulfill all these requirements. The algorithm is implemented within the context of J-Vision, an application for viewing and diagnosing medical images which is currently in use at various hospitals.
[sample image] Mastering Interactive Virtual Bronchioscopy on a Low–End PC (2000)
EndoWeb is a method in the area of virtual endoscopy, that allows a quick and easy hybrid visualization. Overlays of different visualization methods (e.g., surface rendering and volume rendering with different transfer functions) are produced in real time on a low-end PC. To achieve real time frame rates, image-based rendering-techniques have been used.
[sample image] Mastering Windows: Improving Reconstruction (2000)
Ideal reconstruction filters, for function or arbitrary derivative reconstruction, have to be bounded in order to be practicable since they are infinite in their spatial extend. This can be accomplished by multiplying them with windowing functions. In this paper, we discuss and assess the quality of commonly used windows and show that most of them are unsatisfactory in terms of numerical accuracy. Particularly useful are the Kaiser and Gaussian windows since both have a parameter to control the shape of the window, which, on the other hand, requires to find appropriate values for these parameters. We show how to derive optimal parameter values for Kaiser and Gaussian windows using a Taylor series expansion of the convolution sum.
[sample image] Mastering Perspective Projection through Parallelly Projected Slabs for Virtual Endoscopy (2000)
Virtual Endoscopy is a promising medical application for volume rendering techniques where perspective projection is mandatory. Most of the acceleration techniques for direct volume rendering make use of parallel projection. This is also the case of the current generation of VolumePro systems, which achieve real-time frame rates but unfortunately just provide parallel projection. In this paper, an algorithm to approximate perspective volume rendering using parallelly projected slabs is presented. The introduced error due to the approximation is investigated. Based on the error estimation, an improvement to the basic algorithm is presented. The improvement increases the frame rate keeping the global maximal error bounded. The usability of the algorithm is shown through the virtual endoscopic investigation of various types of medical data sets.
[sample image] Mastering Transfer Function Specification by using VolumePro Technology (2000)
A new user-interface paradigm for the specification of transfer functions is presented. The specification is usually a difficult task as mapping information for a number of different domains (data range, color, opacity, etc.) has to be defined. In the presented approach, the definition of the mapping information can be realized independently for each property domain. A set of specification tools is provided for each domain, enabling users with different levels of experience or demanding time restrictions to choose an appropriate approach for their needs. Real-time feedback during the manipulation of parameters has been proven to be crucial to the specification. An interactive direct-volume-rendering display is realized by utilizing dedicated hardware acceleration.
[sample image] Interactive High-Quality Maximum Intensity Projection (2000)
Maximum Intensity Projection (MIP) is a volume rendering technique which is used to visualize high-intensity structures within volumetric data. At each pixel the highest data value, which is encountered along a corresponding viewing ray is depicted. MIP is, for example, commonly used to extract vascular structures from medical data sets (angiography). Due to lack of depth information in MIP images, animation or interactive variation of viewing parameters is frequently used for investigation. Up to now no MIP algorithms exist which are of both interactive speed and high quality. In this paper we present a high-quality MIP algorithm (trilinear interpolation within cells), which is up to 50 times faster than brute-force MIP and at least 20 times faster than comparable optimized techniques. This speed-up is accomplished by using an alternative storage scheme for volume cells (sorted by value) and by removing cells which do not contribute to any MIP projection (regardless of the viewing direction) in a preprocessing step. Also, a fast maximum estimation within cells is used to further speed up the algorithm.
[sample image] Gradient Estimation in Volume Data using 4D Linear Regression (2000)
In this paper a new gradient estimation method is presented which is based on linear regression. Previous contextual shading techniques try to fit an approximate function to a set of surface points in the neighborhood of a given voxel. Therefore, a system of linear equations has to be solved using the computationally expensive Gaussian elimination. In contrast, our method approximates the density function itself in a local neighborhood with a 3D regression hyperplane. This approach also leads to a system of linear equations but we will show that it can be solved with an efficient convolution. Our method provides at each voxel location the normal vector and the translation of the regression hyperplane which are considered as a gradient and a filtered density value respectively. Therefore, this technique can be used for surface smoothing and gradient estimation at the same time.
[sample image]Curvature-Based Transfer Functions for Direct Volume Rendering (2000)
We present a new concept of transfer functions for direct volume rendering. In contrast to previous work, we attempt to define a transfer function in the domain of principal curvature magnitudes. Such a definition helps the user to suppress or enhance structures of a specific shape class. It also allows to set a smooth color or opacity transition within thick surfaces or even solid objects. From the user's point of view the attractiveness of such transfer functions resides in their easy, (semi)automatic specification.
[sample image]The Multi-Dimensional Hartley Transform as a Basis for Volume Rendering (1999)
The Fast Hartley Transform (FHT), a discrete version of the Hartley Transform (HT), has been studied in various papers and shown to be faster and more convenient to implement and handle than the corresponding Fast Fourier Transform (FFT). As the HT is not as nicely separable as the Fourier Transform (FT), a multidimensional version of the HT needs to perform a final correction step to convert the result of separate HTs for each dimension into the final multi-dimensional transform. Although there exist algorithms for two and three dimensions, no generalization to arbitrary dimensions can be found in the literature. We demonstrate an easily comprehensible and efficient implementation of the fast HT and its multi-dimensional extension. By adapting this algorithm to volume rendering by the projection-slice theorem and by the use for filter analysis in frequency domain we further demonstrate the importance of the HT in this application area.
[sample image]AlVis - An Aluminium-Foam Visualization and Investigation Tool (1999)
In recent years there has been an increased interest in metal foams in the field of material science. The stress absorbing potential is one of the most interesting properties for the application of aluminium foam (e.g. car manufacturing). Material scientists need to investigate the structure of metal foams in order to optimize their deformation behavior. An interactive tool for the investigation is presented in this paper.
[sample image] VirEn: A Virtual Endoscopy System (1999)
Virtual endoscopy systems are promising tools for the simplification of daily clinical procedures. In this paper, a conceptual framework for a virtual endoscopy system (VirEn) is proposed, which is intended to be a highly interactive system. Research efforts have concentrated on the generation of an optimal path for the automated navigation of the data set. Extensions to existing thinning algorithms used to generate the optimal path are presented and discussed. First results produced with VirEn are shown.
[sample image] Fast Surface Rendering of Volumetric Data (1999)
In this paper a new direct volume-rendering method is presented for fast display of iso-surfaces. In order to reduce the data to be processed, the algorithm eliminates those voxels which are invisible from a specific domain of viewing directions. The remaining surface points are stored in an appropriate data structure optimized for fast shear-warp projection. The proposed data structure also supports the application of cutting planes in order to visualize the internal part of the volume as well. Unlike many other surface-oriented techniques, the presented method does not trade image quality for speed. It does not require any specialized hardware either to achieve interactive frame rates, thus it can be widely used in medical imaging applications even on low end hardware.
[sample image] Multiple Views and Magic Mirrors - Multimodal Volume Visualization of the Human Brain (1998)
Multimodal visualization of functional and anatomical data of the human brain is an important field in medical volume visualization. The aim of this application is to provide the user with information on the location of functional activations in the different regions of the brain. When the approach of direct volume rendering is chosen, the visual impression of details usually suffers from accumulating the image from colors and opacities derived from the data set. We present extensions to overcome this problem: Transfer function volumes are used for the highlighting of activated regions. Multiple views simplify the task of localizing these parts of the brain. Complementary information is brought into the visualization by Magic Mirrors in order to enhance the comprehensive view of the multimodal arrangement of volume data sets.
[sample image] Real-Time Maximum Intensity Projection (1998)
Maximum Intensity Projection (MIP) is a volume rendering technique which is used to extract high-intensity structures from volumetric data. At each pixel the highest data value encountered along the corresponding viewing ray is determined. MIP is commonly used to extract vascular structures from medical MRI data sets (angiography). The usual way to compensate for the loss of spatial and occlusion information in MIP images is to view the data from different view points by rotating them. As the generation of MIP is usually non-interactive, this is done by calculating multiple images offline and playing them back as an animation. We developed a new algorithm which is capable of interactively generating Maximum Intensity Projection images using parallel projection and templates. Voxels of the data set which will never contribute to a MIP due to their neighborhood are removed during a preprocessing step. The remaining voxels are stored in a way which guarantees optimal cache coherency regardless of the viewing direction. For use on low-end hardware, a preview-mode is included which renders only more significant parts of the volume during user interaction. Furthermore our data structure can be used for extensions of the MIP technique like MIP with depth-shading and Local Maximum Intensity Projection (LMIP).



Institute of Computer Graphics / Visualization and Animation Group / Research

This page is maintained by Andreas König. It was last updated on January 28, 2002.
If you have any comments, please send a message to helwig@cg.tuwien.ac.at.