Level of Detail (LoD) techniques for real-time... In this paper, we are particularly interested in the problem of realistic collision simulation in scenes where large numbers of objects are colliding and processing must occur in real-time. An interruptible and therefore degradable collision handling mechanism is used and the perceptual impact of this degradation is explored. We look for ways in which we can optimise the realism of such simulations and describe a series of psychophysical experiments that investigated different factors affecting collision perception, including eccentricity, separation, distractors, causality and accuracy of physical response. Finally, strategies for incorporating these factors into a perceptually adaptive real-time simulation of large numbers of visually similar objects are presented.

The perceived quality of computer graphics imagery depends on the accuracy of the rendered frames, as well as the capabilities of the human visual system. Fully detailed, high fidelity frames still take many minutes even hours to render on today's computers. The human eye is physically incapable of capturing a moving scene in full detail. We sense image detail only in a 2 # foveal region, relying on rapid eye movements, or saccades, to jump between points of interest. Our brain then reassembles these glimpses into a coherent, but inevitably imperfect, visual percept of the environment. In the process, we literally lose sight of the unimportant details. In this paper, we demonstrate how properties of the human visual system, in particular inattentional blindness, can be exploited to accelerate the rendering of animated sequences by applying a priori knowledge of a viewer's task focus. We show in a controlled experimental setting how human subjects will consistently fail to notice degradations in the quality of image details unrelated to their assigned task, even when these details fall under the viewers' gaze. We then build on these observations to create a perceptual rendering framework that combines predetermined task maps with spatiotemporal contrast sensitivity to guide a progressive animation system which takes full advantage of image-based rendering techniques. We demonstrate this framework with a Radiance ray-tracing implementation that completes its work in a fraction of the normally required time, with few noticeable artifacts for viewers performing the task.

Interactive simulation is made possible in many applications by simplifying or culling the finer details that would make real-time performance impossible. This paper examines detail simplification in the specific problem of collision handling for rigid body animation. We present an automated method for calculating consistent collision response at different levels of detail. The mechanism works closely with a system which uses a pre-computed hierarchical volume model for collision detection. 1.

We present a new algorithm for best-effort simplification of polygonal meshes based on principles of visual perception. Building on previous work, we use a simple model of low-level human vision to estimate the perceptibility of local simplification operations in a view-dependent Multi-Triangulation structure. Our algorithm improves on prior perceptual simplification approaches by accounting for textured models and dynamic lighting effects. We also model more accurately the scale of visual changes resulting from simplification, using parametric texture deviation to bound the size (represented as spatial frequency) of features destroyed, created, or altered by simplifying the mesh. The resulting algorithm displays many desirable properties: it is viewdependent, sensitive to silhouettes, sensitive to underlying texture content, and sensitive to illumination (for example, preserving detail near highlight and shadow boundaries, while aggressively simplifying washed-out regions). Using a unified perceptual model to evaluate these effects automatically accounts for their relative importance and balances between them, overcoming the need for ad hoc or hand-ttmed heuristics.

Level of detail (LOD) is a technique where geometric objects are represented at a number of resolutions, allowing the workload of the system to be modulated on-line. There are numerous schemes for implementing LOD, using selection criteria based upon an object's distance, size, velocity, or eccentricity. However, little is known about how to specify optimally when a particular LOD should be selected so that the user is not aware of any visual change, or to what extent any particular LOD scheme can improve an application 's performance. In response, this paper produces a generic, orthogonal model for LOD based upon data from the field of human visual perception. The effect of this model on the system is evaluated to discover the contribution that each component makes towards any performance improvement. The results suggest that both velocity and eccentricity LOD should be implemented together (if at all) because their individual contribution is likely to be negligible. Also, it is apparent that size (or distance) optimisations offer the greatest benefit, contributing around 95% of any performance increment.

Modern real-time graphics systems are required to render millions of polygons to the screen per second. However, even with this high polygon rendering bandwidth, there are still applications which tax this rendering capability. We introduce in this paper a technique which adaptively allocates polygons to objects in a scene according to their visual importance. It is expected that using this technique, an improvement in the perceptual quality of a rendered image should result, for the same overall number of polygons being rendered. We present both a theoretical basis and a complete design for a visual attention-based level of detail management technique. We also present some preliminary assessment of output from the system. Applications for this technique are expected to be found in the areas of entertainment, visualisation and simulation.

The complexity of most virtual environments prevents them being rendered in real time even on modern graphics hardware. Knowledge of the visual system of the user viewing the environment may be used to significantly reduce image computation times. In this paper, we demonstrate the principle of Change Blindness, a major side effect of brief visual disruptions, including an eye saccade, a flicker, or a blink, where portions of the scene that have changed simultaneously with the visual disruption go unnoticed to the viewer. The onset of the visual disruption inhibits visual attention by swamping the user’s local motion signals, short-circuiting the automatic system that normally draws attention to the change location. Without automatic control, attention is controlled entirely by slower, higher-level mechanisms in the visual system, that search the scene, object by object, until attention finally focuses on the object that is changing. Previous work in perception-based rendering has exploited human visual acuity, to control detail (and therefore time) spent on rendering parts of a scene. In our experiment we show that if changes in rendering detail occur when there is a visual disruption, then visual attention to the change is dramatically slowed as in natural scenes. Therefore, if the principal is used in dynamic animations the change will have passed through the visual field without notice before the viewers ’ attention has picked up the change. Our results clearly show that flaws in the human visual system, such as Change Blindness, can be exploited to reduce rendering times substantially without compromising perceived visual quality.

A major challenge in Virtual Reality is to achieve realism at interactive rates. However, the computational time required for realistic image synthesis is significant, precluding such realism in real-time. This paper demonstrates a concept that may be exploited to reduce rendering times substantially without compromising perceived visual quality in interactive tasks. We demonstrate

Measures of dissimilarity of 3D models are necessary in a wide range of applications such as geometry compression, simplification, and 3D model retrieval. In many cases a metric that models perceptual dissimilarity is desirable. Recently, metrics for 3D models have been evaluated in that respect using concepts such as just noticeable differences, rankings, and others. We propose a simple experimental setup for evaluating supra-threshold perception of 3D models in which users select models at equal perceptual distance to given pairs of models. We discuss the advantages of our approach and report the results of a field study comparing six objective distance measures applied to palettes of simplified reference models. We found that the objective measures are biased, and generally imagebased metrics perform better than metrics based on the original 3D geometry.

Abstract. In virtual environments, head pose and/or eye-gaze estimation can be employed to improve the visual experience of the user by enabling adaptive level of detail during rendering. In this study, we present a real-time system for rendering complex scenes in an immersive virtual environment based on head pose estimation and perceptual level of detail. In our system, the position and orientation of the head are estimated using stereo vision approach and markers placed on a pair of glasses used to view images projected on a stereo display device. The main innovation of our work is the incorporation of uncertainty estimates to improve the visual experience perceived by the user. The estimated pose and its uncertainty are used to determine the desired level of detail for different parts of the scene based on criteria originating from physiological and psychological aspects of human vision. Subject tests have been performed to evaluate our approach. 1