Spatial alignment of cameras
Volumetric video capture with RGB-D cameras

We have previously obtained npoint moving clouds from n cameras, each capturing two streams, RGB and D.

The spatial alignment or "registration" stage consists in finding the position and rotation of the cameras in space, which can generally be stored in a 4x4 matrix. This alignment enables the point clouds to be brought together coherently, the first step in reconstructing the character.

There are several ways of achieving this. From our succession of point clouds, we extract a synchronized image, or frame, for each camera. Although it is possible to interpolate several images to improve quality, we will only present the alignment of a single frame here.