Method and device for identifying and extracting images of multiple users, and for recognizing user gestures

摘要

A method identifies and extracts images of one or more users. The method obtains an array of pixel values that constitutes a scene image and a corresponding array of depth values that constitutes a depth map. The depth map and the image are registered. The method obtains an array containing the 3D positions of the pixel value points in real-world coordinates by coordinate transformation of the depth map and image. The method then clusters the points into groups according to their relative positions so each group contains points in the same region of space and corresponds to a user location. The method defines individual volumes of interest around each user location. The method selects points from the array of 3D positions located in the volumes of interest to obtain segmentation masks for each user. The segmentation masks are then applied to the image to extract images of the users.