Tensor Displays treated a stack of LCDs as a compressive light field

What you see

The paper's own project page, archived from the MIT Media Lab Camera Culture Group's site, frames the result as glasses-free: a viewer standing at different positions in front of a stacked-LCD panel sees different views of a scene, without headset, glasses, or shutter, because the display emits distinct light rays in different directions rather than one flat image. The page's photographs show a benchtop prototype built from modified LCD panels and a custom backlight, not a finished consumer product, and describe wide viewing separation as the visible effect a bystander would notice while walking past it.

How it works

Titled “Tensor Displays: Compressive Light Field Synthesis using Multilayer Displays with Directional Backlighting,” the paper by Gordon Wetzstein, Douglas Lanman, Matthew Hirsch, and Ramesh Raskar — confirmed by its Crossref publication record, whose author list differs from an earlier hint naming Wolfgang Heidrich — describes a family of light-field displays built from a stack of time-multiplexed, light-attenuating layers under uniform or directional backlighting. The paper's abstract states that an N-layer, M-frame version of this stack can be represented mathematically as a tensor, and introduces an optimization method, nonnegative tensor factorization, to compute what each layer should show so the combined result approximates a target light field. This is computational light-field emission through patterned attenuation, not a hologram recorded from interference and not simple stereoscopic 3D.

Viewing conditions and limits

The abstract itself declines to give a specific resolution, contrast ratio, or brightness figure, stating only comparative claims: that the tensor-display architectures it identifies achieve “greater depths of field, wider fields of view, and thinner form factors” than prior automultiscopic displays, verified through simulation and a reconfigurable prototype. Because those are comparisons to earlier lab displays, not absolute specifications, they should not be read as consumer-display numbers. The abstract confirms a GPU-based implementation was fast enough for interactive use, without stating a specific frame rate.

What it is not

This is not an interference-recorded hologram: nothing here uses coherent laser light or captures a wavefront, and the abstract's own vocabulary is light-field synthesis and tensor factorization, not holography. It is also not a shipped television; the described hardware is a lab-built, reconfigurable prototype assembled from modified LCD panels, explicitly a research vehicle for testing the tensor-display architectures the paper proposes, not a product specification.

• Does the source describe a hologram, a light-field display, or ordinary stereoscopic 3D, and which term does the paper itself use?
• Are the reported advantages measured against prior displays in the same family, or against no baseline at all?
• Is the described system a lab prototype or a specified, purchasable product?

Read against its own abstract, the paper's contribution is a unifying mathematical framework for a class of layered, backlit displays, tested on one reconfigurable bench rig rather than delivered as a finished screen, a distinction that later commercial light-field and volumetric products inherit conceptually without inheriting this exact hardware.