Image Blending and View Clustering for Multi-Viewer Immersive Projection Environments

Abstract

Investment into multi-wall Immersive Virtual Environments is often motivated by the potential for small groups of users to work collaboratively, yet most systems only allow for stereographic rendering from a single viewpoint. This paper discusses approaches for supporting copresent head-tracked users in an immersive projection environment, such as the CAVETM, without relying on additional projection and frame-multiplexing technology. The primary technique presented here is called Image Blending and consists of rendering independent views for each head-tracked user to an off-screen buffer and blending the images into a final composite view using view-vector incidence angles as weighting factors. Additionally, users whose view-vectors intersect a projection screen at similar locations are grouped into a view-cluster. Clustered user views are rendered from the average head position and orientation of all users in that cluster. The clustering approach minimizes users' exposure to undesirable display artifacts such as inverted stereo pairs and nonlinear object projections by distributing projection error over all tracked viewers. These techniques have the added advantage that they can be easily integrated into existing systems with minimally increased hardware and software requirements. We compare Image Blending and View Clustering with previously published techniques and discuss possible implementation optimizations and their tradeoffs.