Image Prediction Based on Kalman Filtering for Interactive Simulation Environments

In this paper a new method is presented to increase the frame rate of CGI (computer generated images) applications by means of image prediction. This method is based on Kalman filtering for the prediction of the new viewpoint position. The method processes two consecutive input images and the dedicated z-buffer content, which are computed by the graphics hardware, for the prediction of the next (future) image frame. The input data is transformed into an output image by 3D image warping. An enhanced z-buffer algorithm decides which pixels are visible. Key-Words: Kalman filter, image prediction, interactive simulation