Free-viewpoint photorealistic view synthesis from casually captured video
A sampling density for capturing a plurality of two-dimensional images of a three-dimensional scene may be determined. The sampling density may be below the Nyquist rate. However, the sampling density may be sufficiently high such that captured images may be promoted to multiplane images and used to generate novel viewpoints in a light field reconstruction framework. Recording guidance may be provided at a display screen on a mobile computing device based on the determined sampling density. The recording guidance identifying a plurality of camera poses at which to position a camera to capture images of the three-dimensional scene. A plurality of images captured via the camera based on the recording guidance may be stored on a storage device.
1. A method comprising:
determining via a processor a sampling density for capturing a plurality of two-dimensional images of a three-dimensional scene, wherein the sampling density is below a Nyquist rate, wherein the sampling density is sufficiently high such that the plurality of two-dimensional images may be promoted to multiplane images and used to generate novel viewpoints in a light field reconstruction framework, wherein the sampling density is a function of a designated number of depth planes at which the plurality of two-dimensional images are decomposed;
providing recording guidance at a display screen on a mobile computing device based on the sampling density, the recording guidance identifying a plurality of camera poses at which to position a camera to capture the plurality of two-dimensional images of the three-dimensional scene; and
storing on a storage device the plurality of two-dimensional images captured via the camera based on the recording guidance.
2. The method recited in claim 1 , wherein each of the plurality of two-dimensional images is associated with a virtual frustum, and wherein every point in the three-dimensional scene falls within two or more of the virtual frustums.
3. The method recited in claim 1 , wherein the sampling density is determined as a maximum pixel disparity of any scene point between adjacent two-dimensional images of the plurality of two-dimensional images.
4. The method recited in claim 1 , wherein the sampling density is a function of a minimum depth associated with the three-dimensional scene.
5. The method recited in claim 4 , wherein the minimum depth identifies a minimum distance from a respective camera viewpoint associated with one of the plurality of two-dimensional images and a physical point included in the three-dimensional scene.
6. The method recited in claim 1 , wherein the sampling density is determined based on one or more initial input images of the three-dimensional scene.
7. The method recited in claim 1 , wherein the display screen is configured to present a live camera view, and wherein the recording guidance includes a visual indicator overlaid on the live camera view.
8. The method recited in claim 7 , wherein the visual indicator indicates both a target position and a target orientation of the camera.
9. The method recited in claim 7 , wherein the visual indicator includes a grid of partially transparent objects.
10. The method recited in claim 1 , wherein the mobile computing device is a smartphone.
11. A computing device comprising a processor and a memory storing instructions for execution by the processor to perform a method, the method comprising:
determining via the processor a sampling density for capturing a plurality of two-dimensional images of a three-dimensional scene, wherein the sampling density is below a Nyquist rate, wherein the sampling density is sufficiently high such that the plurality of two-dimensional images may be promoted to multiplane images and used to generate novel viewpoints in a light field reconstruction framework, wherein the sampling density is a function of a designated number of depth planes at which the plurality of two-dimensional images are decomposed;
providing recording guidance at a display screen on a mobile computing device based on the sampling density, the recording guidance identifying a plurality of camera poses at which to position a camera to capture the plurality of two-dimensional images of the three-dimensional scene; and
storing on a storage device the plurality of two-dimensional images captured via the camera based on the recording guidance.
12. The computing device recited in claim 11 , wherein each of the plurality of two-dimensional images is associated with a virtual frustum, and wherein every point in the three-dimensional scene falls within two or more of the virtual frustums.
13. The computing device recited in claim 11 , wherein the sampling density is determined as a maximum pixel disparity of any scene point between adjacent two-dimensional images of the plurality of two-dimensional images.
14. The computing device recited in claim 11 , wherein the sampling density is a function of a minimum depth associated with the three-dimensional scene.
15. The computing device recited in claim 14 , wherein the minimum depth identifies a minimum distance from a respective camera viewpoint associated with one of the plurality of two-dimensional images and a physical point included in the three-dimensional scene.
16. The computing device recited in claim 11 , wherein the sampling density is determined based on one or more initial input images of the three-dimensional scene.
17. The computing device recited in claim 11 , wherein the display screen is configured to present a live camera view, and wherein the recording guidance includes a visual indicator overlaid on the live camera view.
18. One or more non-transitory computer readable media storing instructions for execution by a processor to perform a method, the method comprising:
determining via the processor a sampling density for capturing a plurality of two-dimensional images of a three-dimensional scene, wherein the sampling density is below a Nyquist rate, wherein the sampling density is sufficiently high such that the plurality of two-dimensional images may be promoted to multiplane images and used to generate novel viewpoints in a light field reconstruction framework, wherein the sampling density is a function of a designated number of depth planes at which the plurality of two-dimensional images are decomposed;
providing recording guidance at a display screen on a mobile computing device based on the sampling density, the recording guidance identifying a plurality of camera poses at which to position a camera to capture the plurality of two-dimensional images of the three-dimensional scene; and
storing on a storage device the plurality of two-dimensional images captured via the camera based on the recording guidance.
19. The one or more non-transitory computer readable media recited in claim 18 , wherein each of the plurality of two-dimensional images is associated with a virtual frustum, and wherein every point in the three-dimensional scene falls within two or more of the virtual frustums.
20. The one or more non-transitory computer readable media recited in claim 18 , wherein the sampling density is determined as a maximum pixel disparity of any scene point between adjacent two-dimensional images of the plurality of two-dimensional images.