The Mirror Stereo technology

The human visual system perceives 3D by combining the images received by two eyes. Copying this so-called stereo processing ability of humans has been a research topic within computer vision/artificial intelligence for some time. A typical stereo vision setup for imitating human stereo processing relies on having two cameras facing in the same direction. Such technology is rarely present in today’s mobile devices.

However, nowadays many mobile devices have two cameras, one facing forward and the other backwards. Can we use these two cameras for stereo vision? Yes we can! With the use of our patented technology a ‘dual shot’ image taken in front of a mirror does just this. Because of the mirror, we have two cameras focused on the subject, even though they have a different focal length. With our technology we can ‘reposition’ the cameras and calibrate a stereo setup which can then produce the desired 3D image.

In our 3DefineMe app we reconstruct the profile of a human face. We use standard computer vision methods to find faces and fit a 3D mask model on our reconstructed 3D point cloud. As a result, we get the model of the face. After this step we reconstruct the facial parameters (e.g. the height, width of the nose, the size of the jaw or the distance between the eyes).

For more information on our technology, please read our patent application titled ‘Method and multi-camera portable device for producing stereo images’.

In the 3DefineMe app we add virtual glasses to the 3D models to provide an augmented reality try-on solution for the eyewear industry.