Real-World Virtual Reality With Head-Motion Parallax.

Area of Science:

• Computer Vision
• Virtual Reality Technology
• Human-Computer Interaction

Background:

• Current virtual reality (VR) content often lacks head-motion parallax, leading to visual-vestibular conflict.
• This conflict diminishes the sense of presence and can cause nausea and discomfort for users.
• Head-motion parallax is crucial for a more immersive and comfortable VR experience.

Purpose of the Study:

• To present an end-to-end framework for enabling head-motion parallax in VR for real-world scenes.
• To capture and render novel views that adapt to user head movements.
• To compare two intermediate representations for their effectiveness in VR content.

Main Methods:

• Utilized a camera rig with vertically stacked rings of outward-facing cameras to capture multi-perspective data.
• Developed an offline processing pipeline to create a compact intermediate representation of the scene.
• Rendered novel views for head-mounted displays based on viewer head movements.
• Compared Stacked OmniStereo and Layered Depth Panoramas as intermediate representations.

Main Results:

• The framework successfully enables head-motion parallax for real-world VR content.
• Evaluated objective image quality metrics for the intermediate representations.
• Assessed the occurrence of disocclusion holes in synthesized novel views for both representations.

Conclusions:

• The developed framework provides a viable solution for incorporating head-motion parallax into VR.
• The comparison of intermediate representations offers insights into optimizing VR content delivery.
• Future work can focus on refining these representations for even greater visual fidelity and user comfort.