A paradigm for the design of systems that manage level of detail in virtual environments is proposed. As an example of the prototyping step in this paradigm, a user study was performed to evaluate the effectiveness of high-detail insets used with head-mounted displays. Ten subjects were given a simple search task that required the location and identification of a single target object. All subjects used seven different displays (the independent variable), varying in inset size and peripheral detail, to perform this task Frame rate, target location, subject input method, and order of display use were all controlled. Primary dependent measures were search time on trials with correct identification, and the percentage of all trials correctly identified. ANOVAs of the results showed that insetless, high-detail displays did not lead to significantly different search times or accuracies than displays with insets. In fact, only the insetless, low-detail display returned significantly different results. Further research is being performed to examine the effect of varying task complexity, inset size, and level of detail.

In the companion paper, Watson et al. (1997), we demonstrated the effectiveness of using perceptual criteria to select the amount of detail that is displayed in an immersive virtual reality (VR) system. Based upon this determination, we will now attempt to develop a principled, perceptually oriented framework to automatically select the appropriate level of detail (LOD) for each object in a scene, taking into consideration the limitations of the human visual system. We apply knowledge and theories from the domain of visual perception to the field of VR, thus optimizing the visual information presented to the user based upon solid metrics of human vision. Through a series of contrast grating experiments, a user's visual acuity may be assessed in terms of spatial frequency (c/deg) and contrast. The results of these tests can be modeled mathematically using a contrast sensitivity function (CSF). Therefore, we can use the CSF results to estimate how much visual detail the user can perceive in an object at any instant. Then, if we could describe this object in terms of its spatial frequencies, this would enable us to select the lowest LOD available without the user being able to perceive any visual change.