Visual Perception of Stereo Projections for Virtual Reality

Background

Stereoscopic images are often an essential element of three-dimensional perception in interactive Virtual Reality (VR) systems. In many projector-based VR systems, such as a CAVE or stereowall, two projectors are used to project two seperate images on a screen. By using light filters both in front of the projectors and in the user's glasses, the two images can be perceived as a stereo pair. Together with other depth cues, such as linear perspective and motion parallax, this stereo pair allows the human visual system to reconstruct a mental 3D model of the virtual world.

Problem statement

The perceptual quality of the presented stereo images influences the overall 3D perception. Several optical techniques exist for bringing the computer generated stereo pair to the user through the use of projectors. Given a "correctly" computer generated image, the quality of the resulting percieved image also heavily depend on the components used, such as the filters, screens and projectors. Effects such as screen gain, stereo cross-talk, low brightness and contrast and color effects degrade the resulting images. In practice, we observe that a this degradation of (stereo) image quality influences the ability to effectively and comfortably work on 3D data in VR system.

Task description

In this work, we will investigate the effects of the individual system components on the (stereo) image perception and user performance in 3D tasks. For this, we will look at the interplay between and the relevance of properties such as light intensity, color, viewing angle and stereo separation through entire computer-to-eye chain. A perceptual model will be built which can be calibrated for a certain stereo projector system. The results of the model will be used to compensate or correct the computer generated images, and to again optimize visual perception. By applying this image correction in real-time on the graphics card, our existing VR applications can be optimized for human perception. The effect of the perceptual model on 3D task performance will be evaluated through a set of user studies.

Keywords

Modelling: projector: resolution, focus, DLP image build-up, alignment, synchronisation screen: properties of contrast & brightness, uniformity, gain, polarisation color: perception of intensity, contrast and color resolution stereo: perception of ghosting/crosstalk interaction: moving user, moving eyes, moving content Measuring & Calibration: camera-projector alignment semi-automated model generation & calibration process Real-time image correction on the GPU

Related People, Relevant work:

Projects/VisualPerceptionStereo (last edited 2010-06-21 14:53:00 by localhost)