Apple has registered in the Bureau of patents and trademarks United States patent entitled “Improved face detection using depth information”. It is noteworthy that earlier this technology was owned by the Israeli company PrimeSense, which cupertinos acquired in 2013.
The patent describes software that is able to scan the image for the presence of a human face. This process is much more complicated, if we are not talking about a static image, a video or streaming, as a person can be in frame from different sides and at different distances. The traditional system in such cases, simultaneously start multiple “Windows” of different sizes, which not only increases computational load but also increases the probability of incorrect recognition.
To reduce resource consumption and reduce the probability of error, Apple proposes to add to existing algorithms depth information. These data will be used for intelligent zoom “Windows” in accordance with their depth coordinates on the basis of “standard” size person. The depth map will be drawn up with the help of infrared radiation. A similar system is already in use Microsoft Xbox Kinect sensor.
To date, the patent describes a General solution to recognition, i.e. it can detect that the person is in the frame, but cannot identify who it belongs to. However, this may be the basis of a more complex system of identification of users.
Such a solution can be implemented in the iPhone 8, which will be released this fall. According to the latest forecasts analyst Ming-Chi Kuo, the manufacturer uses the device “revolutionary” front camera, which with the help of infrared sensors will be able to create 3D images and to recognize faces.