This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Eurographics Symposium on Rendering (2006) Tomas Akenine-Mller and Wolfgang Heidrich (Editors) Antialiasing for Automultiscopic 3D Displays M. Zwicker 1 W. Matusik 2 F. Durand 3 H. Pfister 2 1 Department of Computer Science and Engineering, University of California, San Diego 2 Mitsubishi Electric Research Laboratories 3 Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology Abstract Automultiscopic displays show stereoscopic images that can be viewed from any viewpoint without special glasses. They hold great promise for the future of television and digital entertainment. However, the image quality on these 3D displays is currently not sufficient to appeal to the mass market. In this paper, we extend the frequency analysis of light fields to address some of the major issues in 3D cinematography for automultiscopic displays. First, we derive the bandwidth of 3D displays using ray-space analysis, and we introduce a method to quantify display depth of field. We show that this approach provides solid foundations to analyze and distinguish various aspects of aliasing. We then present an anti-aliasing technique for automultiscopic displays by combining a reconstruction and a display prefilter. Next, we show how to reparameterize multi-view inputs to optimally match the depth of field of a display to improve the image quality. Finally, we present guidelines for 3D content acquisition, such as optimal multi-view camera configuration and placement. 1. Introduction For more than a century, the display of three-dimensional images has inspired the imagination and ingenuity of en- gineers and inventors. Automultiscopic displays offer view- ing of high-resolution stereoscopic images from arbitrary positions without glasses. These displays consist of view- dependent pixels that reveal a different color to the ob- server based on the viewing angle. View-dependent pixels can be implemented using conventional high-resolution dis- plays and parallax-barriers (see Figure 1 ), lenticular sheets, or integral lens sheets. Although the optical principles of multiview auto-stereoscopy have been known for over a cen- tury [ Oko76 ], it is only recently that displays with increased resolution have made them practical. Laptops with automul- tiscopic displays are shipping [ Sha05 ], and high-quality au- tomultiscopic desktop monitors cost about $3,000 [ Opt05 ]. As a result, 3D television is getting renewed attention with Grundigs announcement that they will acquire, trans- mit, and display 3D content during the 2006 soccer world cup [ Yah05 ]. However, automultiscopic displays today have several major drawbacks. Most importantly, they are plagued by disturbing visual artifacts that are most salient for moving observers, but also make long viewing from a static posi- tion an uncomfortable experience. Secondly, the acquisition of artifact-free 3D content is challenging. Photographers, videographers, and professionals in the broadcast and movie...
View Full Document
- Spring '08