Live capture, rectification, and streaming of stereoscopic internet video for casual users

We investigate stereoscopic 3D (S3D) video using H.264/AVC supplemental enhancement information messages, real-time rectification, and show objective quality metrics PSNR and MSSIM for depth adjusted and rectified common test sequences converted to side-by-side frame compatible video. We use common computer vision techniques to estimate scene disparity and apply texture warping to vertically align images. The S3D system can capture live video from either two commodity webcams or an inexpensive stereo webcam. Two video streams are captured, rectified, encoded as H.264 video and transmitted to a media server. Clients connect to the media server to receive the H.264 encoded S3D video. A client internet browser-plugin decodes the S3D video as either 2D or S3D if the client has GPU rendering hardware.