| Summary: | 碩士 === 國立臺灣大學 === 資訊工程學研究所 === 90 === Due to the need of more efficient and friendly human computer interface (HCI), studies on face processing have been rapidly expanded in recent years. One useful component for HCI is the autostereoscopic display system. An autostereoscopic display system can provide users great enjoyment of stereo visualization without the uncomfortable and inconvenient drawbacks of wearing stereo glasses or Head-Mounted Displays (HMD). In order to render stereo video with respect to user’s view points and to accurately project stereo video onto the user’s eyes, the left and right eye positions of the user, who is allowed to move around freely, have to be obtained when the user is watching the autostereoscopic display. By first automatically detecting the appearance of user’s frontal face and the tracking the position of the user’s eyes, the autostereoscopic display system can project stereo views precisely to the user’s eyes. In order to acquire the positions of eyes, we have built a real-time eye-tracking system which can automatically detect the position of user’s face and keep tracking the position of the face/eyes in real-time. We propose a method based on PCA (Principal Component Analysis) to detect the frontal faces in images. The face detection module adopt the framework of multi-resolution hierarchy and focus-of-attention. The face candidates are first selected and grouped, then verified to determine whether there exists a face. After the positions of faces in an image are obtained, a fast algorithm for template matching, called the winner-update algorithm is adopted. We utilize the winner-update algorithm to achieve the goal of real-time tracking. After the position of the face is acquired, the positions of the eyes are then located using a convolution-based method according to the geometric relation between the face and the eyes. In addition to be used by the autostereoscopic display system, techniques for face detection and tracking can also be applied to many other applications, such as face recognition and face authentication.
|
|---|
