Method for head image registration and head model generation and corresponding devices

A method for head image registration, comprising:

• providing a plurality of combined 2D and depth images of a head captured from different positions;
• determining landmark points which are predefined points on the head for each of the combined 2D and depth images;
• and
• performing a pairwise coarse registration between pairs of combined 2D and depth images based on the landmark points,
• wherein the method further comprises:
• after the pairwise coarse registration, performing a fine registration of the coarse registered combined 2D and depth images based on full point clouds represented by the combined 2D and depth images,
• wherein the fine registration is performed selectively based on odometry or point to plane ICP (iterative closest point), depending on a geometrical difference between the combined 2D and depth images of the respective pair after the coarse registration, and
• wherein determining the landmark points comprises determining a visibility probability of the landmark points based on a predetermined depth assigned to each landmark point and the 2D images of the combined 2D images;
• selecting landmark points based on the visibility probability;
• and
• providing 3D landmark points for the selected landmark points based on the combined 2D and depth images.

A device, comprising a processor configured to:

• receive a plurality of combined 2D and depth images of a head captured from different positions;
• determine landmark points which are predefined points on the head for each of the combined 2D and depth images;
• and
• perform a pairwise coarse registration between pairs of combined 2D and depth images based on the landmark points,
• wherein the processor is further configured to:
• after the coarse registration, perform a fine registration of the coarse registered combined 2D and depth images based on full point clouds represented by the combined 2D and depth images,
• wherein the fine registration is performed selectively based on odometry or point to plane ICP (iterative closest point), depending on a geometrical difference between the combined 2D and depth images of the respective pair after the coarse registration, and
• wherein determining the landmark points comprises determining a visibility probability of the landmark points based on a predetermined depth assigned to each landmark point and the 2D images of the combined 2D images;
• selecting landmark points based on the visibility probability;
• and
• providing 3D landmark points for the selected landmark points based on the combined 2D and depth images.