Color and infra-red three-dimensional reconstruction using implicit radiance functions
An image is rendered based a neural radiance field (NeRF) volumetric representation of a scene, where the NeRF representation is based on captured frames of video data, each frame including a color image, a widefield IR image, and a plurality of depth IR images of the scene. Each depth IR image is captured when the scene is illuminated by a different pattern of points of IR light, and the illumination by the patterns occurs at different times. The NeRF representation provides a mapping between positions and viewing directions to a color and optical density at each position in the scene, where the color and optical density at each position enables a viewing of the scene from a new perspective, and the NeRF representation provides a mapping between positions and viewing directions to IR values for each of the different patterns of points of IR light from the new perspective.
1 . A method comprising:
receiving an image rendered based on a neural radiance field (NeRF) representation of a scene based on captured frames of video data,
at least one of the captured frames including:
a color image of the scene,
a widefield infrared (IR) image of the scene captured when the scene is illuminated by a source of widefield IR light, and
a plurality of depth IR images of the scene captured when the scene is illuminated by different patterns of points of IR light occurring at different times,
the NeRF representation providing a mapping between positions and viewing directions to a color and optical density at each position in the scene to enable a viewing the scene from a new perspective, and
the NeRF representation providing a mapping between positions and viewing directions to IR values for each of the different patterns of points of IR light from the new perspective;
defining a synthesizing function to generate an enhanced image from the image; and
computing the synthesizing function using a neural network trained based on reducing a loss between a predicted image generated by the neural network and a ground truth image captured by a ground truth camera during training.
2 . The method according to claim 1 , wherein the method further includes, prior to receiving the image:
capturing a 3D model of a scene using a volumetric capture system; and
rendering the image based on the NeRF representation.
3 . The method according to claim 2 , wherein the ground truth camera and the volumetric capture system are both directed to a view during training, the ground truth camera producing higher quality images than the volumetric capture system.
4 . The method according to claim 1 , wherein the NeRF representation of the scene is configured to generate, for a plurality of viewing positions and viewing directions with respect to the scene, a position-specific and viewpoint-specific IR image data of the scene.
5 . The method according to claim 4 , wherein the position-specific and viewpoint-specific IR image data includes widefield IR image data and patterned IR image data for each of the different patterns of points of IR light.
6 . The method according to claim 1 , wherein the loss includes a reconstruction loss based on a reconstruction difference between a segmented ground truth image mapped to activations of layers in a neural network and a segmented predicted image mapped to activations of layers in a neural network, the segmented ground truth image being segmented by a ground truth segmentation mask to remove background pixels and the segmented predicted image being segmented by a predicted segmentation mask to remove background pixels, wherein the predicted segmentation mask is predicted based on a combination of both visible light information captured for a frame and IR light captured for a frame.
7 . The method according to claim 6 , wherein the reconstruction difference is saliency re-weighted to down-weight reconstruction differences for pixels above a maximum error or below a minimum error.
8 . The method according to claim 1 , wherein the loss includes a mask loss based on a mask difference between a ground truth segmentation mask and a predicted segmentation mask, wherein the predicted segmentation mask is predicted based on a combination of both visible light information captured for a frame and IR light captured for a frame.
9 . The method according to claim 1 , wherein the predicted image is one of a predicted stereo pair of images and the loss includes a stereo loss based on a stereo difference between the predicted stereo pair of images.
10 . The method according to claim 1 , wherein the neural network is based on a fully convolutional model.
11 . The method according to claim 1 , wherein the computing the synthesizing function using a neural network comprises:
computing the synthesizing function for a left eye viewpoint; and
computing the synthesizing function for a right eye viewpoint.
12 . The method according to claim 1 , wherein the computing the synthesizing function using a neural network is performed in real time.
13 . The method according to claim 1 , wherein the color image of the scene for a frame includes a de-mosaiced color channel.
14 . The method according to claim 1 , wherein the color image of the scene for a frame includes three independent, non-de-mosaiced color channels, and wherein captured frames of video data upon which the NeRF representation of the scene is based include the three independent, non-de-mosaiced color channels.
15 . A performance capture system comprising:
a volumetric capture system configured to render at least one image, including imperfections, reconstructed from at least one viewpoint of a captured 3D model, the at least one image being based on a neural radiance field (NeRF) representation; and
a rendering system configured to receive the at least one image from the volumetric capture system and to generate, in real time, at least one enhanced image in which the imperfections of the at least one image are reduced, the rendering system including a neural network configured to generate the at least one enhanced image by training prior to use, the training including reducing a loss between predicted images generated by the neural network during training and corresponding ground truth images captured by at least one ground truth camera coordinated with the volumetric capture system during training, wherein the reducing of the loss is based on captured frames of video data,
at least one of the each-captured frames including:
a color image of a scene,
a widefield infrared (IR) image of the scene captured when the scene is illuminated by a source of widefield IR light, and
a plurality of depth IR images of the scene captured when the scene is illuminated by different patterns of points of IR light,
the NeRF representation providing a mapping between positions and viewing directions to IR values for each of the different patterns of points of IR light from a new perspective.
16 . The performance capture system according to claim 15 , wherein the at least one ground truth camera is included in the performance capture system during training and otherwise not included in the performance capture system.
17 . The performance capture system according to claim 15 , wherein the volumetric capture system includes a single active stereo camera directed to a single view and, during training, includes a single ground truth camera directed to the single view.
18 . The performance capture system according to claim 15 , wherein the volumetric capture system includes a plurality of active stereo cameras directed to multiple views and, during training, includes a plurality of ground truth cameras directed to the multiple views.
19 . The performance capture system according to claim 15 , wherein the performance capture system includes a stereo display configured to display one of the at least one enhanced image as a left eye view and one of the at least one enhanced image as a right eye view.
20 . The performance capture system according to claim 19 , wherein the performance capture system includes a telepresence communication node.
21 . The performance capture system according to claim 19 , wherein the stereo display is included in an augmented reality (AR) headset.
22 . The performance capture system according to claim 19 , wherein the stereo display is a head-tracked auto-stereo display.