Tom Vierjahn | Westphalian University of Applied Science (original) (raw)

Papers by Tom Vierjahn

2018 IEEE Conference on Virtual Reality and 3D User Interfaces (VR)

2017 IEEE 3rd Workshop on Everyday Virtual Reality (WEVR)

2016 IEEE 6th Symposium on Large Data Analysis and Visualization (LDAV), 2016

2016 Third Workshop on Visual Performance Analysis (VPA), 2016

: Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned... more : Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned level of detail. For each step the number of triangles was halved until reaching 312 triangles (far right).

: Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned... more : Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned level of detail. For each step the number of triangles was halved until reaching 312 triangles (far right).

: Iterative reconstruction of a textured triangle mesh. From left to right: The point cloud creat... more : Iterative reconstruction of a textured triangle mesh. From left to right: The point cloud created from a set of seven aerial images, and the triangle mesh after 0.04 s, 0.2 s and 1 s of reconstruction. The grey proxy is provided as a visual 3D-cue.

In this work we present an improved rendering algorithm for the planning of stage lighting using ... more In this work we present an improved rendering algorithm for the planning of stage lighting using photon mapping, since existing software implements algorithms that do not suit the lighting workflow in a theatre adequately. During theatre lighting, positional changes of light sources are occuring rarely compared to intensity and color changes: Once a light is positioned and aimed correctly, many intensity and color values are tested. Thus, intensity and color changes must be recomputed very fast whereas positional changes may take longer to be rendered. The algorithm we present provides such fast recalculations and furthermore balances idle times of user and renderer.

ABSTRACT Modern interactive environments are tending to not impose any restrictions by additional... more ABSTRACT Modern interactive environments are tending to not impose any restrictions by additional encumbrance on the user. Therefore it seems reasonable to utilize computer vision for detecting interactions. There are many techniques available to segment an image into fore- and background, but most of them require the scene to be static. In contrast this work applies to installations where an interactive projection is in the field of vision of a camera. Within the bounds of this work an existing technique utilizing infra-red light, which has been shown to be illumination-variant, was to be improved. First several techniques for foreground-background-segmentation were checked for suitability. The most appropriate was to be refined by adding depth segmentation to suppress the animated background utilizing stereo-vision. This turned out to be impractical for the given installations, but more practicable ones were found. In addition to this even the favoured segmentationtechnique may provide an improvement on the existing one.

In this work a preprocessing step is presented which allows for segmentation of objects located i... more In this work a preprocessing step is presented which allows for segmentation of objects located in front of a projected, animated -i.e. dynamic -background. The described technique adapts and improves a stereoscopic approach which makes use of the depth constraint of the dynamic background to pre-compute an accurate remapping of a pair of camera images. The difference of the remapped images indicates foreground objects due to slight differences of the remapping of their pixels since these objects do not satisfy the depth constraint. That way, the dynamic background can be easily eliminated in a succeeding foreground-background-segmentation step. The approach is experimentally analyzed with respect to accuracy and minimal foregroundbackground distance. In contrast to computationally expensive real-time depth segmentation this approach is well suited for rather low priced marketing or point of sale installations.

Computers & Graphics, 2015

ABSTRACT In this paper we propose surface-reconstructing growing neural gas (SGNG), a learning ba... more ABSTRACT In this paper we propose surface-reconstructing growing neural gas (SGNG), a learning based artificial neural network that iteratively constructs a triangle mesh from a set of sample points lying on an object's surface. From these input points (SGNG) automatically approximates the shape and the topology of the original surface. It furthermore assigns suitable textures to the triangles if images of the surface are available that are registered to the points. By expressing topological neighborhood via triangles, and by learning visibility from the input data, SGNG constructs a triangle mesh entirely during online learning and does not need any post-processing to close untriangulated holes or to assign suitable textures without occlusion artifacts. Thus, SGNG is well suited for long-running applications that require an iterative pipeline where scanning, reconstruction and visualization are executed in parallel. Results indicate that SGNG improves upon its predecessors and achieves similar or even better performance in terms of smaller reconstruction errors and better reconstruction quality than existing state-of-the-art reconstruction algorithms. If the input points are updated repeatedly during reconstruction, SGNG performs even faster than existing techniques.

3D-Längenanamorphosen sind verzerrte 2D-Bilder, die sich von einer bestimm- ten Perspektive betra... more 3D-Längenanamorphosen sind verzerrte 2D-Bilder, die sich von einer bestimm- ten Perspektive betrachtet scheinbar zu einem 3D-Objekt verwandeln. Durch die Compu- tergrafik ist es ein Leichtes geworden, 3D-Anamorphosen zu erzeugen. Doch das derzeitig angewandte Verfahren bietet noch viel Raum für Optimierungen. So werden im Moment noch zwei verlustbehaftete Rendervorgänge zur Erzeugung der 3D-Längenanamorophose be- nötigt. Diese Arbeit legt den Grundstein für ein neues, schnelleres und qualitativ besseres Verfahren zur Erzeugung von 3D-Anamorphosen mittels gängiger 3D-Programme. Durch geeignete Vorverzerrung der 3D-Szene kann die Anamorphose mit nur einem verlustfreien Rendervorgang erzeugt werden. Auch die Integration in Echtzeit-Anwendungen ist dank der einfachen Berechnung möglich. Kombiniert mit einem Personen-Tracking-System sind so zum Beispiel auch interaktive Augmented-Reality-Anwendungen möglich.

2011 IEEE International Symposium on VR Innovation, 2011

In this paper we introduce a novel approach to interactively generate a virtual environment (VE) ... more In this paper we introduce a novel approach to interactively generate a virtual environment (VE) as reconstruction of the real world by using a swarm of autonomously flying miniature unmanned aerial vehicles (MUAVs). In our system each of the MUAVs is equipped with advanced network as well as different sensor technologies, which allow to capture images from low altitudes and

2018 IEEE Conference on Virtual Reality and 3D User Interfaces (VR)

2017 IEEE 3rd Workshop on Everyday Virtual Reality (WEVR)

2016 IEEE 6th Symposium on Large Data Analysis and Visualization (LDAV), 2016

2016 Third Workshop on Visual Performance Analysis (VPA), 2016

: Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned... more : Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned level of detail. For each step the number of triangles was halved until reaching 312 triangles (far right).

: Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned... more : Reconstruction of Stanford Bunny with 20,000 triangles (far left) and the automatically learned level of detail. For each step the number of triangles was halved until reaching 312 triangles (far right).

: Iterative reconstruction of a textured triangle mesh. From left to right: The point cloud creat... more : Iterative reconstruction of a textured triangle mesh. From left to right: The point cloud created from a set of seven aerial images, and the triangle mesh after 0.04 s, 0.2 s and 1 s of reconstruction. The grey proxy is provided as a visual 3D-cue.

In this work we present an improved rendering algorithm for the planning of stage lighting using ... more In this work we present an improved rendering algorithm for the planning of stage lighting using photon mapping, since existing software implements algorithms that do not suit the lighting workflow in a theatre adequately. During theatre lighting, positional changes of light sources are occuring rarely compared to intensity and color changes: Once a light is positioned and aimed correctly, many intensity and color values are tested. Thus, intensity and color changes must be recomputed very fast whereas positional changes may take longer to be rendered. The algorithm we present provides such fast recalculations and furthermore balances idle times of user and renderer.

ABSTRACT Modern interactive environments are tending to not impose any restrictions by additional... more ABSTRACT Modern interactive environments are tending to not impose any restrictions by additional encumbrance on the user. Therefore it seems reasonable to utilize computer vision for detecting interactions. There are many techniques available to segment an image into fore- and background, but most of them require the scene to be static. In contrast this work applies to installations where an interactive projection is in the field of vision of a camera. Within the bounds of this work an existing technique utilizing infra-red light, which has been shown to be illumination-variant, was to be improved. First several techniques for foreground-background-segmentation were checked for suitability. The most appropriate was to be refined by adding depth segmentation to suppress the animated background utilizing stereo-vision. This turned out to be impractical for the given installations, but more practicable ones were found. In addition to this even the favoured segmentationtechnique may provide an improvement on the existing one.

In this work a preprocessing step is presented which allows for segmentation of objects located i... more In this work a preprocessing step is presented which allows for segmentation of objects located in front of a projected, animated -i.e. dynamic -background. The described technique adapts and improves a stereoscopic approach which makes use of the depth constraint of the dynamic background to pre-compute an accurate remapping of a pair of camera images. The difference of the remapped images indicates foreground objects due to slight differences of the remapping of their pixels since these objects do not satisfy the depth constraint. That way, the dynamic background can be easily eliminated in a succeeding foreground-background-segmentation step. The approach is experimentally analyzed with respect to accuracy and minimal foregroundbackground distance. In contrast to computationally expensive real-time depth segmentation this approach is well suited for rather low priced marketing or point of sale installations.

Computers & Graphics, 2015

ABSTRACT In this paper we propose surface-reconstructing growing neural gas (SGNG), a learning ba... more ABSTRACT In this paper we propose surface-reconstructing growing neural gas (SGNG), a learning based artificial neural network that iteratively constructs a triangle mesh from a set of sample points lying on an object's surface. From these input points (SGNG) automatically approximates the shape and the topology of the original surface. It furthermore assigns suitable textures to the triangles if images of the surface are available that are registered to the points. By expressing topological neighborhood via triangles, and by learning visibility from the input data, SGNG constructs a triangle mesh entirely during online learning and does not need any post-processing to close untriangulated holes or to assign suitable textures without occlusion artifacts. Thus, SGNG is well suited for long-running applications that require an iterative pipeline where scanning, reconstruction and visualization are executed in parallel. Results indicate that SGNG improves upon its predecessors and achieves similar or even better performance in terms of smaller reconstruction errors and better reconstruction quality than existing state-of-the-art reconstruction algorithms. If the input points are updated repeatedly during reconstruction, SGNG performs even faster than existing techniques.

3D-Längenanamorphosen sind verzerrte 2D-Bilder, die sich von einer bestimm- ten Perspektive betra... more 3D-Längenanamorphosen sind verzerrte 2D-Bilder, die sich von einer bestimm- ten Perspektive betrachtet scheinbar zu einem 3D-Objekt verwandeln. Durch die Compu- tergrafik ist es ein Leichtes geworden, 3D-Anamorphosen zu erzeugen. Doch das derzeitig angewandte Verfahren bietet noch viel Raum für Optimierungen. So werden im Moment noch zwei verlustbehaftete Rendervorgänge zur Erzeugung der 3D-Längenanamorophose be- nötigt. Diese Arbeit legt den Grundstein für ein neues, schnelleres und qualitativ besseres Verfahren zur Erzeugung von 3D-Anamorphosen mittels gängiger 3D-Programme. Durch geeignete Vorverzerrung der 3D-Szene kann die Anamorphose mit nur einem verlustfreien Rendervorgang erzeugt werden. Auch die Integration in Echtzeit-Anwendungen ist dank der einfachen Berechnung möglich. Kombiniert mit einem Personen-Tracking-System sind so zum Beispiel auch interaktive Augmented-Reality-Anwendungen möglich.

2011 IEEE International Symposium on VR Innovation, 2011

In this paper we introduce a novel approach to interactively generate a virtual environment (VE) ... more In this paper we introduce a novel approach to interactively generate a virtual environment (VE) as reconstruction of the real world by using a swarm of autonomously flying miniature unmanned aerial vehicles (MUAVs). In our system each of the MUAVs is equipped with advanced network as well as different sensor technologies, which allow to capture images from low altitudes and