The relative contributions of stereo, lighting, and background scenes in promoting 3D depth visualization

Reviewer: Patrick Gilles Maillot

One of the major problems in 3D scene visualization has always been how to ensure a correct interpretation of the data, that is, how to ensure that people looking at rendered scenes will correctly interpret the relative positions and sizes of the objects constituting the scenes. That is only half the problem: a bigger challenge is to enable people to use a scene to enter data with correct position, elevation, and size relative to elements already in the scene. The authors present a well-rounded study of the relative contribution of different cues that help users judge the depth, elevation, priority, size, and position of elements in a rendered 3D scene. In the introduction, the authors recall the tremendous progress in computer performance, which has enabled both the proliferation of 3D systems and the use of better rendering techniques. Yet it is still a challenge to offer a display that is sufficiently unambiguous that all users will interpret the data in exactly the same way. The study starts with binocular disparity and the use of stereoscopic glasses and rendering systems. Other cues are worth mentioning and studying. The authors base their study on cue theory, which raises the question of whether different stimulus dimensions are perceived as separate elements, or if they are contributors to a single object. Several models have been proposed, and they often conflict or contradict each other. Some models argue that different cues follow a “weighted additive” rule; some promote the idea of more complex combinations and are called “multiplicative”; some mention a “vetoing” mechanism that strongly advantages the strongest cue; and finally there are “fusion” models, where cues interact in either a linear or a nonlinear way. The authors recognize some cues as more dominant than others. Stereopsis and movement are certainly in this category. Shadows also contribute to our perception of the shape and orientation of objects. Several studies have shown that humans intrinsically perceive unseen light sources as coming from above, which is usually valid in a terrestrial environment. Shadows are also helpful in perceiving object size, elevation, and depth. The authors conducted a study, focusing on shadows and stereo viewing, whose conditions included shadows and shadow casting, viewing mode (mono or stereo), and scene background. They assembled a representative panel of users and measured the ability of each to position, resize, and manipulate objects. Each part of the study took into account the accuracy and response time of each experiment, and the findings are discussed and analyzed. An almost obvious conclusion is that stereo vision is more powerful than other cues, such as shadows, but a more interesting finding is that response time worsens when shadow cues are added to stereo vision, which is clearly in opposition to the additive or multiplicative cue theories. The authors present their conclusions in a format that will be useful to 3D user interface designers. They also propose further research, to increase our understanding of the effects of shadows on human perception and our ability to perform spatial tasks. This paper presents constructive results about 3D computer user interfaces and opens an interesting area of research—the impact of shadows on human 3D perception.