The tracking 3D display systems by image processing analysis

This paper describes a fast algorithm of the face detection for an autostereoscopic display system allowing the viewing zone to follow the observer's head. This two phases face detection algorithm may real time detect the observer's eye-position such that the observers don't need to wear any artificial instruments to enable a most natural tracking 3D display system. The two phases face detection algorithm correlates the input image and an eigenface first, then send this correlated result through two thresholdings, and a filter to generate a possible face region. Then, the combination of the possible face region and the correlated result generates the face position information. Clustered background and eye-glasses wearing is allowed. This algorithm is also relatively robust on scaling and viewing angle variations.