Generating Virtual Images for Multi-view Video

As for the research in the Free view television (FTV) and stereo video areas, the accurate multi-view video capture is a expensive precondition. We present a virtual multi-view video capture system based on OpenGL programming. By using this system, a researcher may create arbitrary 3D scenes, and capture images with depth maps by setting up multiple virtual cameras, whose intrinsic and extrinsic parameters can be imported from or exported to the outside.We conducted several experiments within it to perform camera calibration, depth estimation and view-points synthesis. The system has been demonstrated to be a powerful tool for evaluation of stereo video algorithms.