A light field is a claim about how many rays you capture

The 1996 light field rendering paper defines the function a light-field display must approximate, and what capturing it costs.

Historical event

August 1, 1996

First source published

August 1, 1996

Site publication

September 18, 2026

A 1996 definition, not a marketing term

'Light-field display' gets used loosely, so it is worth returning to where the term comes from. In August 1996, Marc Levoy and Pat Hanrahan presented Light Field Rendering at SIGGRAPH, defining the light field as 'a 4D function that completely characterizes the flow of light through unobstructed space in a static scene with fixed illumination'. That is a precise claim: not a picture, not a stereo pair, but a function assigning a radiance value to every ray passing through a region. The ACM proceedings record confirms the paper ran in the main SIGGRAPH 1996 programme, pages 31 to 42.

What it takes to build and view one

The paper's method was to assemble arrays of images, computer-rendered or captured with a video camera on a motorised gantry, and treat the resulting set as samples of the 4D function. New viewpoints then come 'simply by combining and resampling the available images' rather than by reconstructing scene geometry. For a display rather than a rendering algorithm, the implication is blunt: reproducing a light field means emitting a distinct ray of light for every direction the function specifies at every point on the display surface, not just showing one image or even a small stereo pair. A viewer moving their head, or several viewers at different seats, should each intercept a different, geometrically correct subset of rays. The paper reports compression of the captured data 'exceeding 100:1 with very little loss of fidelity', a figure about data storage, not about how many physical rays a screen can actually produce.

Why the label gets stretched

Because the original definition concerns ray count and sampling density, not any particular hardware, a product calling itself a 'light-field display' is making an implicit claim about how finely it samples that 4D function; a lenticular panel with a handful of views is a coarse approximation, and a true dense-ray system is a much harder engineering problem. Neither document opened here evaluates any commercial display against that bar; they establish only the definition and the rendering method the term originates from. This is an editorial distinction worth insisting on: whether a given product's ray density is close to or far from what the paper implies is a separate, product-specific question this record does not settle.

Questions to bring to a demo

• How many distinct rays or views does the display emit per resolvable angle, and over what field of view?
• Does the image change correctly for horizontal and vertical head movement, or only one axis?
• Is the source content an actual captured or rendered light field, or a stereo pair relabelled?

Light field rendering set the vocabulary in 1996 before any display could match it. Judging a 'light-field' product means asking how much of that 4D function it actually reproduces, not whether it uses the phrase.

Sources & reading trail

States the paper's core definition of a light field as the 4D function describing radiance through unobstructed space, and describes generating new views by resampling it without depth or feature matching.

Source published: 1 August 1996 · Retrieved: 16 September 2026

Confirms the paper's formal publication details: title, authors, SIGGRAPH 1996 proceedings, and page range 31 to 42.

Source published: 1 August 1996 · Retrieved: 16 September 2026

Primary documents establish the record; the mechanism reading and the demo questions are Presence Atlas editorial analysis. This retrospective draft does not imply the site published on the event date.

Continue reading

• A patented 2001 screen design lets a room share one 3D view
• A modular panel's own patents specify a 120-degree zone
• A lens array let a headset show where the eye should focus
• Browse the complete the archive

Published September 18, 2026, not on the date of the event described.