Digital Signal Processing for Enhanced Automultiscopic 3-D Visualization


Voronoi cell of reciprocal lattice modeling the layout of single-view pixels in a 3-D lenticular display

Team: P. Agniel, A. Jain, J. Konrad
Funding: National Science Foundation (ITR)
Status: Completed (2002-2006)

Background: Next-generation display systems will include depth in order to invoke a “being there” experience. This can be achieved by presenting two separate perspectives to a viewer by means of a 3-D stereoscopic display (anaglyph, polarized, shuttered, autostereoscopic). Such displays, however, suffer from two deficiencies: the need to use glasses (except for autostereoscopic displays)  and/or absence of motion parallax (change of perspective with viewer motion). In the last few years, a new generation of advanced automultiscopic (multiview, no glasses) 3-D displays has been introduced on the market. Although pioneers in this area, such as Stereographics Corp. and Philips Electronics NV, have withdrawn from the market, others such as Newsight GmbH and Dimension Technologies Inc. continue to deliver 3-D screens with no glasses. Such displays require multiple views (up to 9) thus putting significant strain on transmission. Furthermore, is individual views are not suitably band-limited, the resulting 3-D image suffers from spatial aliasing.

Summary: In this project,  we developed optimal multiplexing of views in order to invoke as natural 3-D perception as possible. One of the issues is anti-alias pre-filtering of each view. Since individual views are sampled irregularly during multiplexing, we use 2-D lattice approximations and derive anti-alias filter specifications. Application of such filters results in reduced aliasing and improved 3-D perception. We also studied extensions of lattice-modeled pre-filters to the case of filters modeled based on unions of cosets. We also investigated optimal pre-filter specifications based on Markov-1 AR image model while accounting for crosstalk from neighboring views.