The present disclosure provides a distance determination method, apparatus and system, relating to the technical field of image processing. The method includes the following steps: acquiring a master visual image photographed by a master camera and an original auxiliary visual image photographed by an auxiliary camera; acquiring an initial matching point pair between the master visual image and the original auxiliary visual image through feature extraction and feature matching; correcting the original auxiliary visual image sequentially, based on the initial matching point pair and different constraints, so as to obtain a target auxiliary visual image, wherein the different constraints includes: a constraint of a minimum rotation angle and a constraint of a minimum parallax; and determining a focusing distance according to the master visual image and the target auxiliary visual image. The focusing distance can be determined more accurately.

The present disclosure provides a distance determination method, apparatus and system, relating to the technical field of image processing. The method includes the following steps: acquiring a master visual image photographed by a master camera and an original auxiliary visual image photographed by an auxiliary camera; acquiring an initial matching point pair between the master visual image and the original auxiliary visual image through feature extraction and feature matching; correcting the original auxiliary visual image sequentially, based on the initial matching point pair and different constraints, so as to obtain a target auxiliary visual image, wherein the different constraints includes: a constraint of a minimum rotation angle and a constraint of a minimum parallax; and determining a focusing distance according to the master visual image and the target auxiliary visual image. The focusing distance can be determined more accurately.

An image-based method of modeling and rendering a three-dimensional model of an object is provided. The method comprises: obtaining a three-dimensional point cloud at each frame of a synchronized, multi-view video of an object, wherein the video comprises a plurality of frames; extracting a feature descriptor for each point in the point cloud for the plurality of frames without storing the feature descriptor for each frame; producing a two-dimensional feature map for a target camera; and using an anti-aliased convolutional neural network to decode the feature map into an image and a foreground mask.

An image-based method of modeling and rendering a three-dimensional model of an object is provided. The method comprises: obtaining a three-dimensional point cloud at each frame of a synchronized, multi-view video of an object, wherein the video comprises a plurality of frames; extracting a feature descriptor for each point in the point cloud for the plurality of frames without storing the feature descriptor for each frame; producing a two-dimensional feature map for a target camera; and using an anti-aliased convolutional neural network to decode the feature map into an image and a foreground mask.

A method for training a depth estimation model with depth hints is disclosed. For each image pair: for a first image, a depth prediction is determined by the depth estimation model and a depth hint is obtained; the second image is projected onto the first image once to generate a synthetic frame based on the depth prediction and again to generate a hinted synthetic frame based on the depth hint; a primary loss is calculated with the synthetic frame; a hinted loss is calculated with the hinted synthetic frame; and an overall loss is calculated for the image pair based on a per-pixel determination of whether the primary loss or the hinted loss is smaller, wherein if the hinted loss is smaller than the primary loss, then the overall loss includes the primary loss and a supervised depth loss between depth prediction and depth hint. The depth estimation model is trained by minimizing the overall losses for the image pairs.

A method for training a depth estimation model with depth hints is disclosed. For each image pair: for a first image, a depth prediction is determined by the depth estimation model and a depth hint is obtained; the second image is projected onto the first image once to generate a synthetic frame based on the depth prediction and again to generate a hinted synthetic frame based on the depth hint; a primary loss is calculated with the synthetic frame; a hinted loss is calculated with the hinted synthetic frame; and an overall loss is calculated for the image pair based on a per-pixel determination of whether the primary loss or the hinted loss is smaller, wherein if the hinted loss is smaller than the primary loss, then the overall loss includes the primary loss and a supervised depth loss between depth prediction and depth hint. The depth estimation model is trained by minimizing the overall losses for the image pairs.

A first aspect of the invention provides a method of training a neural network for capturing volumetric video, comprising: generating a 3D model of a scene; using the 3D model to generate a high fidelity depth map; capturing a perceived depth map of the scene, having a field of view that is aligned with a field of view of the high fidelity depth map; and training the neural network based on the high fidelity depth map and the perceived depth map, wherein the high fidelity depth map has a higher fidelity to the scene than the perceived depth map has.

A first aspect of the invention provides a method of training a neural network for capturing volumetric video, comprising: generating a 3D model of a scene; using the 3D model to generate a high fidelity depth map; capturing a perceived depth map of the scene, having a field of view that is aligned with a field of view of the high fidelity depth map; and training the neural network based on the high fidelity depth map and the perceived depth map, wherein the high fidelity depth map has a higher fidelity to the scene than the perceived depth map has.

Systems and methods for hybrid depth regularization in accordance with various embodiments of the invention are disclosed. In one embodiment of the invention, a depth sensing system comprises a plurality of cameras; a processor; and a memory containing an image processing application. The image processing application may direct the processor to obtain image data for a plurality of images from multiple viewpoints, the image data comprising a reference image and at least one alternate view image; generate a raw depth map using a first depth estimation process, and a confidence map; and generate a regularized depth map. The regularized depth map may be generated by computing a secondary depth map using a second different depth estimation process; and computing a composite depth map by selecting depth estimates from the raw depth map and the secondary depth map based on the confidence map.

Systems and methods for hybrid depth regularization in accordance with various embodiments of the invention are disclosed. In one embodiment of the invention, a depth sensing system comprises a plurality of cameras; a processor; and a memory containing an image processing application. The image processing application may direct the processor to obtain image data for a plurality of images from multiple viewpoints, the image data comprising a reference image and at least one alternate view image; generate a raw depth map using a first depth estimation process, and a confidence map; and generate a regularized depth map. The regularized depth map may be generated by computing a secondary depth map using a second different depth estimation process; and computing a composite depth map by selecting depth estimates from the raw depth map and the secondary depth map based on the confidence map.