Generalized Scene Reconstruction

Generalized Scene Reconstruction (GSR)

Generalized Scene Reconstruction (GSR) is method of 3D reconstruction from multiple images wherein a scene model representing a generalized light field and a relightable matter field is created.

– Popular Definition from Wikipedia

Scene reconstruction methods like 3D scanning, photogrammetry, or LiDAR only reconstruct the matter field, and paint the light field onto surfaces as color. By contrast, GSR represents all the light moving through all points in a scene in any direction (the plenoptic light field), as well as the way materials interact with that light.

GSR can reconstruct the kinds of objects and scenes that normally create stumbling blocks for scene reconstruction (such as transparent materials, finely structured geometry, or complex lighting conditions). Other methods need a highly controlled set of lighting conditions, materials, and topologies in order to function. GSR does not. It can reconstruct any scene on Earth. Quidient proclaims GSR to be the Final Frontier of Photography because it represents the world the same way that humans perceive it – as a set of lighting and objects, independent of one another.

Generalized Scene. A region in space that contains virtually any type of light or material that people encounter in the course of their daily lives. Generalized scenes include objects that are shiny, partially transmissive, featureless, and finely structured. Modern kitchens, contemporary offices, and flower gardens are examples of generalized scenes.

2D Photographs, 3D Scanning, 3D+ or 4D or 5D Generalized Scene Reconstruction with Light Field and Material Properties

Light Field Physics

Light transport theory describes how a given material emits, absorbs, reflects, scatters or transmits various frequencies of light. Quidient decouples the light field from the matter field by using transport theory to represent materials such as metal, wood, glass, and even fog. This decoupling leads to significant advantages in machine learning / AI as well as scene compression and processing speeds.

surface element animation
Plenoptic Database Revolving Image Including Light Field and Matter Field

Spatial (5D) Database

Quidient’s Plenoptic (5D) Database technology represents a major shift in underlying architecture.  It provides a novel means for separately encoding a matter field (as 3D Voxels, shown in turquoise) coincident with any light field (as 2D solid-angle elements, Saels, shown in yellow). This spatially sorted, hierarchical approach leads to randomly accessible and searchable scenes with exceptionally fast subscene insertion and extraction. This is a critical requirement for representing scenes with virtually unlimited levels of detail such as an interactive city map.

Conventional & Polarimetric Imaging

Quidient’s engines are enabled to work with both conventional and polarimetric cameras. Traditional monochrome and color sensors capture the visible color and brightness of a light wave interacting with a material. Polarimetric sensors capture additional information, the light wave’s non-visible “rotational orientation.” When available, this additionally captured information provides important advantages in GSR, especially when dealing with featureless surfaces and low lighting conditions. Quidient’s solution is highly configurable and enabled to use a broad range of individual and/or combination of sensors including today’s newest polarimetric cameras to optimize the scene reconstruction process.

polarimetric imaging figure

Let’s Change the World Together

The notion of “Changing the World” is not a cliché for us. Generalized Scene Reconstruction is a pioneering approach to 3D imaging. All of us on Quidient’s team are optimistic that thousands of GSR-enabled applications will soon transform the way people live and work. We have key positions open for brilliant scientists, digital developers and entrepreneurs.

Sound interesting? Check out our Careers page.


Q icon 77 by 77 pixels
GSR White Paper

Q icon 77 by 77 pixels

Q icon 77 by 77 pixels
Blog - Sizing Hail Dents using Machine Learning

Q icon 77 by 77 pixels
Sign up below to receive an advanced copy of our new GSR White Paper.