state-of-the-art rendering techniques in real-time...

1
State-of-the-art Rendering Techniques in real-time Architectural Visualization SubmissionID: sap 0427 Robert Kuchar * Graphical-Interactive Systems, University of T¨ ubingen Timo Schairer Graphical-Interactive Systems, University of T¨ ubingen Figure 1: Screenshots of different scenes lit and rendered by our system. 1 Introduction Today, high dynamic range (HDR) imaging, image-based lighting (IBL) and special effects techniques are common features found in many non-real-time rendering systems (“offline renderers”). Al- though slowly being adopted by game developers, few real-time rendering tools provide a seamless integration and persistent use of these techniques throughout their systems. In this poster, we present the integration of the aforementioned techniques into a real-time walk-through rendering software based on global illumination used mainly for architectural visualization. Since current graphics boards support floating-point framebuffers, physically based lighting information is preserved throughout the rendering pipeline to allow for photorealistic special effects that are applied in real-time. 2 Lighting When visualizing an architectural scene realistic lighting is crucial in terms of a realistic perceptual experience. Important factors are physical accuracy, authenticity of the light sources and the applied light distribution technique. We cover these requirements by com- bining a physically based daylight model and image based lighting using HDR light probes with progressive refinement radiosity. Regarding the fact, that scenes from architectural visualization can be very detailed and the calculation time is proportional to the num- ber of samples per patch, effective optimizations techniques must be applied. We use importance driven sampling techniques [Ostro- moukhov et al. 2004] and early sample rejection mechanisms that are especially tailored for architectural scenes, typically including both, indoor and outdoor environments. 3 Real-time Rendering The other essential aspect in creating a photorealistic visualization is the integration of effects that compensate the limited capabilities * e-mail: [email protected] e-mail: [email protected] of common display devices. We increase the level of depth per- ception by simulating the depth of field effect [Scheuermann 2004] in conjunction with automatic focus that is essential in the walk- through context. The main limitation when rendering HDR scenes is the insufficient dynamic range of the display device, implying a tone mapping to displayable brightness values. To solve this prob- lem, we extended Reinhard’s photographic global operator [Rein- hard et al. 2002] to respond dynamically to the amount of light in the current view as the user is moving through the virtual scene. To partly recreate the brightness impression of the HDR scene we implemented a glare simulation that is performed before the tone mapping process. All these viewpoint dependent effects are calcu- lated using image processing methods and algorithms. To guarantee realtime performance, we take advantage of the enormous comput- ing power of current graphics hardware by applying GPGPU tech- niques in our post-processing pipeline. To further enhance the visual quality of architectural visualizations, a realistic simulation of surface materials is important. Therefore we provide techniques like parallax bump mapping, fresnel prop- erty, blurred dynamic HDR reflections and first-order refractions in our visualization system. 4 Conclusion We presented a photorealistic rendering system for static scenes that can rival many “off-line renderers” in terms of quality while achiev- ing real-time frame rates. References OSTROMOUKHOV, V., DONOHUE, C., AND J ODOIN, P.-M. 2004. Fast hierarchical importance sampling with blue noise proper- ties. ACM Trans. Graph. 23, 3, 488–495. REINHARD, E., STARK, M., SHIRLEY, P., AND FERWERDA, J. 2002. Photographic tone reproduction for digital images. ACM Trans. Graph. 21, 3, 267–276. SCHEUERMANN, T. 2004. Advanced depth of field.

Upload: buidieu

Post on 01-Aug-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

State-of-the-art Rendering Techniques in real-time Architectural VisualizationSubmissionID: sap 0427

Robert Kuchar∗

Graphical-Interactive Systems, University of TubingenTimo Schairer†

Graphical-Interactive Systems, University of Tubingen

Figure 1: Screenshots of different scenes lit and rendered by our system.

1 Introduction

Today, high dynamic range (HDR) imaging, image-based lighting(IBL) and special effects techniques are common features found inmany non-real-time rendering systems (“offline renderers”). Al-though slowly being adopted by game developers, few real-timerendering tools provide a seamless integration and persistent use ofthese techniques throughout their systems.

In this poster, we present the integration of the aforementionedtechniques into a real-time walk-through rendering software basedon global illumination used mainly for architectural visualization.Since current graphics boards support floating-point framebuffers,physically based lighting information is preserved throughout therendering pipeline to allow for photorealistic special effects that areapplied in real-time.

2 Lighting

When visualizing an architectural scene realistic lighting is crucialin terms of a realistic perceptual experience. Important factors arephysical accuracy, authenticity of the light sources and the appliedlight distribution technique. We cover these requirements by com-bining a physically based daylight model and image based lightingusing HDR light probes with progressive refinement radiosity.

Regarding the fact, that scenes from architectural visualization canbe very detailed and the calculation time is proportional to the num-ber of samples per patch, effective optimizations techniques mustbe applied. We use importance driven sampling techniques [Ostro-moukhov et al. 2004] and early sample rejection mechanisms thatare especially tailored for architectural scenes, typically includingboth, indoor and outdoor environments.

3 Real-time Rendering

The other essential aspect in creating a photorealistic visualizationis the integration of effects that compensate the limited capabilities

∗e-mail: [email protected]†e-mail: [email protected]

of common display devices. We increase the level of depth per-ception by simulating the depth of field effect [Scheuermann 2004]in conjunction with automatic focus that is essential in the walk-through context. The main limitation when rendering HDR scenesis the insufficient dynamic range of the display device, implying atone mapping to displayable brightness values. To solve this prob-lem, we extended Reinhard’s photographic global operator [Rein-hard et al. 2002] to respond dynamically to the amount of light inthe current view as the user is moving through the virtual scene.To partly recreate the brightness impression of the HDR scene weimplemented a glare simulation that is performed before the tonemapping process. All these viewpoint dependent effects are calcu-lated using image processing methods and algorithms. To guaranteerealtime performance, we take advantage of the enormous comput-ing power of current graphics hardware by applying GPGPU tech-niques in our post-processing pipeline.

To further enhance the visual quality of architectural visualizations,a realistic simulation of surface materials is important. Thereforewe provide techniques like parallax bump mapping, fresnel prop-erty, blurred dynamic HDR reflections and first-order refractions inour visualization system.

4 Conclusion

We presented a photorealistic rendering system for static scenes thatcan rival many “off-line renderers” in terms of quality while achiev-ing real-time frame rates.

References

OSTROMOUKHOV, V., DONOHUE, C., AND JODOIN, P.-M. 2004.Fast hierarchical importance sampling with blue noise proper-ties. ACM Trans. Graph. 23, 3, 488–495.

REINHARD, E., STARK, M., SHIRLEY, P., AND FERWERDA, J.2002. Photographic tone reproduction for digital images. ACMTrans. Graph. 21, 3, 267–276.

SCHEUERMANN, T. 2004. Advanced depth of field.