290 likes | 367 Views
A Framework for Perceptual Studies in Photorealistic Augmented Reality. Martin Knecht 1 , Andreas Dünser 2 , Christoph Traxler 1 , Michael Wimmer 1 and Raphael Grasset 3. 1 Institute of Computer Graphics and Algorithms Vienna University of Technology. 2 HIT Lab NZ University of Canterbury.
E N D
A Framework for Perceptual Studies in Photorealistic Augmented Reality Martin Knecht1, Andreas Dünser2,Christoph Traxler1, Michael Wimmer1 and Raphael Grasset3 1 Institute of Computer Graphicsand Algorithms Vienna University of Technology 2 HIT Lab NZ University of Canterbury 3 HIT Lab NZ / ICG University of Canterbury /Graz University of Technology
Introduction • Motivation • Perceptual Issues • Framework • Pilot Study • Limitations • Future Work • Conclusion Martin Knecht
Motivation • Is this a photograph or computer generated? • What visual factors affect your decision? Rademacher P., et al. (2001) Courtesy of Rademacher et al. (2001) Martin Knecht
Motivation • People are very good in judging if an image looks photo-realistic or notBut process behind it not fully understood! • In augmented reality (AR) virtual objects should be rendered in a photorealistic way • Goal: A framework that allows us to study photorealistic rendering techniques at real-time frame rates Martin Knecht
Perceptual Issues • Kruijff E., et al. (2010) distinguish between two types of problems: • Technological limitations • Perceptual nature • Proposed framework focuses on perceptual issues Martin Knecht
Perceptual Pipeline • Perceptual pipeline used to classify issuesKruijff E., et al. (2010) Environment Capturing AR Framework Augmentation Display Device User Martin Knecht
Capturing / Augmentation Stages • Capturing: Covers the process of converting the optical image to a digital one Kruijff E., et al. (2010) • Image resolution, lens distortion,exposure, color correctness,… • Augmentation: Deals with adding the virtual objects into the augmented sceneKruijff E., et al. (2010) • Registration errors, occlusion, rendering,… Martin Knecht
Framework • Our contribution: • Framework to study photorealistic rendering techniques • Interactive tasks with global illumination • Three different rendering modes • Fast prototyping of experiments (XML) • A pilot study to test framework Martin Knecht
Framework • Differential instant radiosity is used to render the scenes Knecht M., et al. (2010) • Rendering modes can be changed during run-time • Mode A: shadows / indirect illumination • Mode B: shadows / indirect illumination • Mode C: shadows / indirect illumination Martin Knecht
Rendering Mode A • Different rendering modes are supported • No shadows • No indirect illumination Martin Knecht
Rendering Mode B • Different rendering modes are supported • With shadows • No indirect illumination Martin Knecht
Rendering Mode C • Different rendering modes are supported • With shadows • With indirect illumination Martin Knecht
Pilot Study • Pilot study to evaluate our framework • Twenty-one participants (15 male, 6 female) • Age between 19 – 59 • 5 tasks Martin Knecht
Pilot Study • Task 1: Estimate distance between real and virtual cubes Martin Knecht
Pilot Study • Task 2: Place real cube at position of virtual one Martin Knecht
Pilot Study • Task 3: Place virtual cube at position of real one using computer keyboard Martin Knecht
Pilot Study • Task 4: Grab and lift real cube Martin Knecht
Pilot Study • Task 5: Grab and lift virtual cube - real cube was used for tactile feedback Martin Knecht
Pilot Study • Duration: 30 to 60 minutes • Interview followed afterwards • Measured • Distance error for tasks 1, 2, 3 • Time for tasks 2, 3, 4, 5 • Analyzed data using non-parametric Friedman tests • Result: No significant effect of rendering technique found Martin Knecht
Pilot Study Martin Knecht
Pilot Study Martin Knecht
Pilot Study • Distance estimation perceived easier by 6 participants in left/right and height direction than in depth • Although no significant effect was found Martin Knecht
Pilot Study • Participants made heavy use of occlusion cue to place cubes in Task 2 & 3 – (19 out of 20) Martin Knecht
Pilot Study • 7 participants preferred manipulation with a computer keyboard (Task 2 vs. Task 3). Martin Knecht
Limitations • No support for stereo rendering • No distance information of environment • Only video see-through HMDs are supported • Tone-mapping still needs manual fine tuning • No support for mobile devices Martin Knecht
Future Work • Enhance rendering framework • Support for stereo rendering • Add camera artifacts • New studies with new tasks • Tasks without occlusion information • Use chin rest to reduce movement • Design new tasks Martin Knecht
Conclusion • We proposed a research test-bed for perceptual studies – still work in progress • Different rendering modes allow new kind of experiments • We showed the results of a pilot study using our framework Martin Knecht
Thank you for your attention! • What features would you like to have in such a framework? Martin Knecht