Kinect-based 3D video conference system

In this paper, we proposed a 3D video conference system based on Kinect. In the sending module, a parallel scheme is designed for data acquisition, preprocessing and encoding of color video frames, depth video frames and audio data. A simple color-matching method is introduced for depth optimization to recover the lost depth information. H.264 and MP3 encoder are used to perform data compression. All the encoded data is transmitted to the receiver using the RTP network protocol. The receiving module is responsible for the decoding of video and audio, 3D rendering and video display. We use Depth-Image-Based-Rendering (DIBR) technology to construct one virtual view scene. Then 3D video is generated using the virtual view video and the decompressed color video. Experimental results show that our system can achieve 15f/s real time communication with better 3D quality.