A Video Encoding System With Conditional Picture-Element Replenishment

Most known methods for efficiently transmitting pictures over communication circuits exploit point-to-point correlation along a scanning line. In particular, point-to-point predictive quantizing systems have been successful, but it is known that there is more correlation between television picture elements in the frame-to-frame time dimension than there is between adjacent elements in a single frame. This is especially true when using the head-and-shoulder view of a person as the picture source, such as is likely to be encountered in a visual communication system. Now that devices for storing large amounts of data and integrated circuits for complex digital processing are available, it is not only possible to take advantage of this fact in picture coding, but it is also economically attractive. In this paper, we describe one method of encoding television signals which takes advantage of the frame-to-frame correlation to reduce the transmission bandwidth. We also describe the experimental facility and the results of initial experiments. "We want to emphasize that the picture source for these experiments is a head-and-shoulder view of a person carrying on an animated 2545