Portrait Vision Fusion for Augmented Reality

摘要

Video communication is a common way to communicate via interactive technology, especially using webcams for remote interaction and for each participant to see each other's characteristics from the screen display. In this paper, the main goal is to augment some dynamic interactive virtual environments. Towards this goal, a method using superimposing a segmented human portrait on a panoramic background is proposed, then the limb interactive element is added into these videos involved with a dynamic portrait segmentation method meanwhile using a Kinect (+openCV) device to extract a portrait for amendment, finally acquires a full portrait of information. Because the face is the most important identification region in a portrait, a head skeleton tracking method is also used to strengthen the remedy for its head segmentation, further uses the edge transparent processing to synthesize them into the video. The approach leds the users can verbally and physically communicate through these video interactive much more vibrantly.