Granted Patent B2

US 12,563,283 · App. 18/318,545 · Granted Feb 24, 2026

Inventors: Murtaza Safdari (Stanford, CA); Sanha Cheong (Stanford, CA); Josef Frisch (Stanford, CA); Sean Gasiorowski (Stanford, CA); Michael Kagan (Stanford, CA); Ariel Schwartzman (Stanford, CA); Maxime Vandegar (Stanford, CA)

Assignee: The Board of Trustees of the Leland Stanford Junior University

H04N23/55G06T15/506H04N23/957

Abstract

Claims (45)

1. A light field imaging system comprising:

a mirror array comprising a plurality of mirrors, where each mirror in the plurality of mirrors is positioned to reflect light from a target object towards a lens such that a different view of the target object is visible in each mirror in the plurality of mirrors from the viewpoint of the lens;

a fixing device configured to position the target object at a fixed distance between the mirror array and the lens;

the lens configured to focus the reflected light;

a sensor configured to capture an image depicting a plurality of reflections of the target object from the mirror array, where the sensor and the mirror array are on opposite sides of the lens;

a processor; and

a memory containing a light field imaging application that directs the processor to:

obtain the captured image;

extract a light field of the target object based on the plurality of reflections of the target object; and

construct a three-dimensional model of the target object based on the light field.

2. The system of claim 1 , wherein to extract the light field of the target object, the light field imaging application further directs the processor to:

map each of the plurality of reflected lights to its corresponding predetermined angle; and

measure an intensity and a direction of each of the plurality of reflected lights.

3. The system of claim 1 , where each view in the plurality of views of the target object is a folded view comprising reflected light rays off of the target object off a corresponding mirror, and light rays directly off of the target object.

4. The system of claim 1 , where the processor constructs the image of the target object by representing the target object as densities dependent on 3D spatial coordinates.

5. The system of claim 1 , where the processor uses a fully differentiable reconstruction model to construct the image of the target object.

6. The system of claim 1 , where constructing the image of the target object is gradient-based.

7. The system of claim 1 , where the mirror array further comprises an alignment grid capable of putting the mirror array in focus of the lens.

8. The system of claim 1 , where the plurality of mirrors comprises a plurality of 5 mm diameter smooth mirrors.

9. The system of claim 1 , where the target object is suspended in front of the center of the mirror array.

10. The system of claim 5 , where the reconstruction model is trained by:

extract a set of views from the plurality of views where each view is a set of pixels and an intensity for each pixel from the set of pixels; and

train the reconstruction model with the extracted set of views.

11. A method for light field imaging, the method comprising:

projecting light at a target object;

reflecting light from the target object towards a lens using mirror array comprising a plurality of mirrors such that a different view of the target object is visible in each mirror in the plurality of mirrors from the viewpoint of the lens;

a fixing device configured to position the target object at a fixed distance between the mirror array and the lens;

focusing the reflected light at the lens;

capturing, using a sensor, an image depicting a plurality of reflections of the target object from the mirror array, where the sensor and the mirror array are on opposite sides of the lens;

obtaining the captured image;

extracting a light field of the target object based on the plurality of reflections of the target object; and

constructing a three-dimensional model of the target object based on the light field.

12. The method of claim 11 , where the extracting of the light field comprises:

mapping each of the plurality of reflected lights to its corresponding predetermined angle; and

measuring an intensity and a direction of each of the plurality of reflected lights.

13. The method of claim 11 , where each view in the plurality of views of the target object is a folded view comprising reflected light rays off of the target object off a corresponding mirror, and light rays directly off of the target object.

14. The method of claim 11 , further comprising constructing the image of the target object by representing the target object as densities dependent on 3D spatial coordinates.

15. The method of claim 11 , where a fully differentiable reconstruction model is used to construct the image of the target object.

16. The method of claim 11 , where the constructing of the image of the target object is gradient-based.

17. The method of claim 11 , where the mirror array is put in focus of the lens by an alignment grid.

18. The method of claim 11 , where the plurality of mirrors comprises a plurality of 5 mm diameter smooth mirrors.

19. The method of claim 11 , where the target object is suspended in front of the center of the mirror array.

20. The method of claim 15 , where the reconstruction model is trained by:

extracting a set of views from the plurality of views where each view is a set of pixels and an intensity for each pixel from the set of pixels; and

training the reconstruction model with the extracted set of views.

Assignments (2)

Continuity (2)

References Cited (20)