Slide1: Interactive Multi-Perspective Views of Virtual 3D Landscape and City Models Haik Lorenz, Matthias Trapp, Markus Jobst, Jürgen Döllner Hasso-Plattner-Institute Computer Graphics Systems Group Prof. Dr. Jürgen Döllner University Potsdam www.hpi.uni-potsdam.de/3d www.3dgi.de


Motivation: Why Multi-Perspective Views ?: Motivation: Why Multi-Perspective Views ? WRAP UP: Goals for 3D Visualization Maintain advantages of 3D visualization Offer navigation and orientation aid Increase the effectiveness of available screen space Reduce noise and dead values in the distance


Existing Solution: Detail + Overview Visualization: Existing Solution: Detail + Overview Visualization Detail Overview


Our Solution: Multi-Perspective Views: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 4 Our Solution: Multi-Perspective Views


Outline: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 5 Outline Related Work Concept Implementation Sketch Performance Results & Discussion Future Work & Open Issues Conclusions


Related Work: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 6 Related Work Art of H.C. Berann Panorama Maps with Non-linear Ray-tracing [Falk ’07] Detail-In-Context Visualization for Satellite Imaginary [Böttger, EG’08]


Concept – Effective Presentation of Spatial 3D Environments : 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 7 Concept – Effective Presentation of Spatial 3D Environments One concept: The Bendable Ground Plate Two 3D visualization approaches: Bird’s Eye View Deformation (Progressive Perspective)  “Which direction am I looking to?” Pedestrian View Deformation (Degressive Perspective)  “Where am I going to?”


Concept – Bird’s eye View Deformation – Parameterization : 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 8 Concept – Bird’s eye View Deformation – Parameterization View-dependent parameters C – camera position β – viewing angle of the reference plane bi – line separating focus and transition zone in the image ri – line of the horizon in the image ri bi


Concept – Pedestrian’s View Deformation – Parameterization: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 9 Concept – Pedestrian’s View Deformation – Parameterization View-dependent parameters C – camera position β – angle between T and T’ db – distance between CT (C projected onto T ) and b ds – width of the transition zone’s source area


Concept – … In Terms of Focus + Context Visualization: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 10 Concept – … In Terms of Focus + Context Visualization


Concept – Graphical Representations of Focus & Context Areas: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 11 Concept – Graphical Representations of Focus & Context Areas Cartographic design of the visualization: Distinct rendering styles for focus and context areas Transition zone: blending between focus and context Vertex-based interpolation (style interpolator)


Implementation Sketch: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 12 Implementation Sketch Clue: Deformation before standard perspective projection Global Deformation [Baar ’84]: Per vertex, GPU based Using vertex shader functionality de Casteljau algorithm for Bézier spline: Characteristics: Single-pass rendering technique Interactive, deformation recalculated per frame No caching of deformed data necessary


Performance Results & Discussion (1): 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 13 Performance Results & Discussion (1) Test Data: 16,000 generically textured buildings, ~100 landmarks, 3 GB color aerial photo, 250 MB grayscale map, Digital terrain model


Performance Results & Discussion (2): 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 14 Performance Results & Discussion (2) Performance Issues: Main bottleneck = texture access Cache efficiency is reduced dramatically Additional data handling overhead for rendering 2 styles Conclusion: Using only adapted view frustum culling is not sufficient  Use hardware-based occlusion culling algorithms  Use distance-based geometric Level of Detail (LoD)


Future Work & Open Issues (1): Future Work & Open Issues (1) Improve User Interaction: Transitions between bird’s eye and pedestrian view deformation Adjust parameterization with respect to users speed or similar Conduct user studies Technical Enhancements: Use dynamic mesh refinement for geometry [Lorenz, WSCG 08] Add thematic information Transfer Concept To: Pedestrian View: Mobile Devices ? Bird’s Eye View: Navigation Systems ?


Future Work & Open Issues (2): 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 16 Future Work & Open Issues (2) [Jobst, VISUAL 08, in review] Focus on cartographic aspects: Minimize transition zone Incorporate levels of detail/abstraction Use non-photorealistic rendering (NPR) [Glander, ACMGIS 2007]


Future Work & Open Issues (3): 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 17 Future Work & Open Issues (3) Silhouette enhancement via geometric scaling


Conclusions: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 18 Conclusions Features: Interactive combination of two views Seamlessly style interpolation Increase effectiveness of representations Extensible concept Performance round-up: Pedestrians View: minimal increase Bird’s Eye View: heavy decrease Possibilities for optimization Future work: Generalization of bendable plane concept More styles for focus & context areas Focus on: the user and more use cases


Slide19: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 19 Demo


Contact: 08.05.2007 Matthias Trapp, Hasso-Plattner-Institut, University Potsdam 20 Contact Thank You ! QUESTIONS ? Matthias Trapp matthias.trapp@hpi.uni-potsdam.de Computer Graphics Systems Group Prof. Dr. Jürgen Döllner www.hpi.uni-potsdam.de/3d Researchgroup 3D-Geoinformation www.3dgi.de