AUTO-Stereoscopic3D Video Conversation System Based On An Improved Eye Tracking Method

The communication system used for Video transmission based on network has been widely applied in daily life. Because of people’s need for videos with higher quality and higher immersion, the Auto-stereoscopic3D Video Conversation System has been a hot spot direction which has attracted attentions from companies and academics. Auto-stereoscopic3D Video Conversation is also the direction of future development of video communications systems.As the streaming media technology and stereoscopic display technology continues to evolve and mature, currently stereoscopic video transmission through network has been widely used in many aspects, such as three-dimensional digital TV, stereo video conferencing systems, video surveillance, telemedicine, distance education etc. This paper implements an auto-stereoscopic3D Video Conversation System. The Sending side captures the reasonable three-dimensional video information using2cameras, encodes the captured stereoscopic video in real time and sends the data using the RTP/RTCP protocol; The Receiving side decodes the video after receiving, then the auto-stereoscopic displayer displays the stereoscopic image according to the observer’s position.To achieve a high degree of freedom during the auto-stereoscopic display, this paper combines the high precision eye-tracking system with a precision lenticular lens array. In the implementation process, to locate human eye position with high accuracy, we combine the Haar-like with the Principal Component Analysis. According to the detected eyes’ positions, the viewing zone in the space will be adjusted by the partially pixilated masks. In this way, the correct stereoscopic effect can be acquired.The entire system is implemented on the VS2010platform, C++is the implementation language. And the OpenCV face database functions are called to achieve human eyes’tracking.