Apple was awarded a patent today that details a method of detecting faces in a digital video feed through the use of depth information.
Published by the U.S. Patent and Trademark Office, the document describes how face detection algorithms could identify the presence of faces in a live video when people in the scene are located at different distances from the camera.
To reduce processing overhead and minimize error rates, the system applies depth information to existing face detection algorithms used in photography and intelligently scales the face window sizes according to their depth coordinates – i.e., the further away a face is from the lens, the smaller the capture frame around it.
The method utilizes a special infrared light to project an optical radiation pattern onto the scene, which is then converted into a depth map. As noted by AppleInsider, the depth mapping system referenced in the patent is based on motion tracking technology developed by Israeli motion capture firm PrimeSense, which Apple acquired in 2013.