A. James Stewart

Fast computation of shadow boundaries using spatial coherence and backprojections

This paper describes a fast, practical algorithm to compute the shadow boundaries in a polyhedral scene illuminated by a polygonal light source. The shadow boundaries divide the faces of the scene into regions such that the structure or “aspect” of the visible area of the light source is constant within each region. The paper also describes a fast, practical algorithm to compute the structure of the visible light source in each region. Both algorithms exploit spatial coherence and are the most efficient yet developed. Given the structure of the visible light source in a region, queries of the form “What specific areas of the light source are visible?” can be answered almost instantly from any point in the region. This speeds up by several orders of magnitude the accurate computation of first level diffuse reflections due to an area light source. Furthermore, the shadow boundaries form a good initial decomposition of the scene for global illumination computations.

General Calculations using Graphics Hardware with Applications to Interactive Caustics

Graphics hardware has been developed with image production in mind, but current hardware can be exploited for much more general computation. This paper shows that graphics hardware can perform general calculations, which accelerate the rendering process much earlier than at the latter image generation stages. An example is given of the real time calculation of refractive caustics.

Hierarchical Visibility in Terrains

This paper describes a hierarchical visibility technique that significantly accelerates terrain rendering. With this technique, large parts of the terrain that are hidden from the viewpoint are culled, thus avoiding the expense of uselessly sending them down the graphics pipeline (only to find in the z—buffer step that they are hidden). The hierarchical visibility technique has been implemented in a multiresolution terrain rendering algorithm and experimental results show very large speedups in some situations.

Effective compression techniques for precomputed visibility

In rendering large models, it is important to identify the small subset of primitives that is visible from a given viewpoint. One approach is to partition the viewpoint space into viewpoint cells, and then precompute a visibility table which explicitly records for each viewpoint cell whether or not each primitive is potentially visible. We propose two algorithms for compressing such visibility tables in order to produce compact and natural descriptions of potentially-visible sets. Alternatively, the algorithms can be thought of as techniques for clustering cells and clustering primitives according to visibility criteria. The algorithms are tested on three types of scenes which have very different structures: a terrain model, a building model, and a world consisting of curved tunnels. The results show that the natural structure of each type of scene can automatically be exploited to achieve a compact representation of potentially visible sets.

ECSGlasses and EyePliances: using attention to open sociable windows of interaction

We present ECSGlasses: wearable eye contact sensing glasses that detect human eye contact. ECSGlasses report eye contact to digital devices, appliances and EyePliances in the users attention space. Devices use this attentional cue to engage in a more sociable process of turn taking with users. This has the potential to reduce inappropriate intrusions, and limit their disruptiveness. We describe new prototype systems, including the Attentive Messaging Service (AMS), the Attentive Hit Counter, the first person attentive camcorder eyeBlog, and an updated Attentive Cell Phone. We also discuss the potential of these devices to open new windows of interaction using attention as a communication modality. Further, we present a novel signal-encoding scheme to uniquely identify EyePliances and users wearing ECSGlasses in multiparty scenarios.

A Uniform Sky Illumination Model to Enhance Shading of Terrain and Urban Areas

Users of geographic information systems (GIS) usually render terrain using a point light source defined by an illumination vector. A terrain shaded from a single point provides good perceptual cues to surface orientation. This type of hill shading, however, does not include any visual cues to the relative height of surface elements. We propose shading the terrain under uniform diffuse illumination, where light arrives equally from all directions of a theoretical sky surrounding the terrain. Surface elements at lower elevations tend to have more of the sky obscured from view and are thus shaded darker. This tinting approach has the advantage that it provides more detailed renderings than point source illumination. We describe two techniques of computing terrain shading under uniform diffuse illumination. One technique uses a GIS–based hill-shading and shadowing tool to combine many point source renderings into one approximating the terrain under uniform diffuse illumination. The second technique uses a C++ computer algorithm for computing the inclination to the horizon in all azimuth directions at all points of the terrain. These virtual horizons are used to map sky brightness to the rendering of the terrain. To evaluate our techniques, we use two Digital Elevation Models (DEMs)—of the Schell Creek Range of eastern Nevada and a portion of downtown Houston, Texas, developed from Light Detection and Ranging (lidar) data. Renderings based on the uniform diffuse illumination model show more detailed changes in shading than renderings based on a point source illumination model.

Using registration uncertainty visualization in a user study of a simple surgical task

We present a novel method to visualize registration uncertainty and a simple study to motivate the use of uncertainty visualization in computer-assisted surgery. Our visualization method resulted in a statistically significant reduction in the number of attempts required to localize a target, and a statistically significant reduction in the number of targets that our subjects failed to localize. Most notably, our work addresses the existence of uncertainty in guidance and offers a first step towards helping surgeons make informed decisions in the presence of imperfect data.

Incremental update of the visibility map as seen by a moving viewpoint in two dimensions

Consider the following problem: A viewpoint moves amongst a set of line segments in the plane and it is desired to maintain the sequence of lines visible from the viewpoint at every increment in its position. The sequence of visible lines is identical for most increments in the position of the viewpoint. It is different only when the viewpoint crosses a visual discontinuity line. Our objective is to be able to quickly report whether the sequence of visible lines needs to be updated and perform the update quickly in that case. We propose an algorithm that satisfies both criteria while using space linear in the number of visual discontinuity lines. This last condition is important because constructing the arrangement of these lines would take space quadratic in their number.

Maintenance of the set of segments visible from a moving viewpoint in two dimensions

Consider a viewpoint moving amongst a set of non– intersecting line segments in the plane. We would like to compute efficiently the set of segments visible at successive positions along the viewpoint trajectory. This problem can be solved in either the off–line or the on–line setting. The latter case, in which the updates of the viewpoint are known only as they occur, is more interesting for interactive systems. This video illustrates a simple algorithm to solve this problem. For a scene containing m visual discontinuities, each discontinuity crossing is performed in time O(log2 m) using space O(m). The space bound is an important consideration, as m can become very large in real applications. Pocchiola [Poc90] maintains each update in the view in logarithmic time for a viewpoint moving amidst a set of objects in the plane (off–line version). For n objects, his algorithm computes the arrangement of 2n curves, which takes space quadratic in n. Bern et al. [BDEG90] address a three– dimensional version of the problem where the trajectory is a straight–line segment and forms part of the input. They show that it is possible in time O((n2 + p) logn) (where p depends on the scene) to find the locations along the segment where the topology of the visible surfaces changes.

LOCAL ROBUSTNESS AND ITS APPLICATION TO POLYHEDRAL INTERSECTION

The field of solid modeling makes extensive ve use of a variety of geometric algorithms. When implemented on a computer, these algorithms often fail because the computer only provides floating point arithmetic, while the algorithms are expecting infinite precision arithmetic on real numbers. These algorithms are not robust. This paper presents a formal theory of robustness. It is then argued that the elegant theoretical approach to robustness is not viable in practice: algorithms like those used in solid modeling are generally too complex for this approach. This paper presents a practical alternative to the formal theory of robustness; this alternative is called local robustness. Local robustness is applied to the design of a polyhedral intersection algorithm, which is an important component in many solid modelers. The intersection algorithm has been implemented, and, in extensive tests, has never failed to produce a valid polyhedron of intersection. A concise characterization of the locally robust intersection algorithm is presented; this characterization can be used to develop variants of the intersection algorithm, and to develop robust versions of other solid modeling algorithms.