Image prediction using face detection and triangulation

摘要

We proposed a new image prediction method that can be applied to the video coding for video-telephony and video- conferencing. Instead of dividing the image frames into 16 × 16 blocks, the frames are triangulated according to a set of control points and edge segments. The control points and the edge segments of the triangular blocks are located at the edge features in the image. These edge segments in the sequence are tracked and are used to estimate the motion and distortion of the triangular blocks in the predicted images. Because we can locate more edge segments on the detected face region than on the background, important facial expression can be better preserved in the predicted images. Our experimental result shows that our method is effective in predicting the image sequence with facial expression.